Body composition predicts hypertension using machine learning methods: A Cohort Study

Body composition predicts hypertension using machine learning methods: A Cohort Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Body composition predicts hypertension using machine learning methods: A Cohort Study Mohammad Ali Nematollahi, Soodeh Jahangiri, Arefeh Asadollahi, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-2232998/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 27 Apr, 2023 Read the published version in Scientific Reports → Version 1 posted 8 You are reading this latest preprint version Abstract Introduction: We used machine learning methods to investigate if body composition indices predict hypertension. Methods: Data from a cohort study was used, and 4663 records were included (2156 were male, 1099 with hypertension, with the age range of 35-70 years old). Body composition analysis was done using bioelectrical impedance analysis (BIA); weight, basal metabolic rate, total and regional fat percentage (FATP), and total and regional fat-free mass (FFM) were measured. We used machine learning methods such as Support Vector Classifier, Decision Tree, Stochastic Gradient Descend Classifier, Logistic Regression, Gaussian Naïve Bayes, K-Nearest Neighbor, Multi-Layer Perceptron, Random Forest, Gradient Boosting, Histogram-based Gradient Boosting, Bagging, Extra Tree, Ada Boost, Voting and Stacking to classify the investigated cases and find the most relevant features to hypertension. Results: FATP, AFFM, BMR, FFM, TRFFM, AFATP, LFATP, and older age were the top features in hypertension prediction. Arm FFM, basal metabolic rate, total FFM, Trunk FFM, leg FFM, and male gender were inversely associated with hypertension, but total FATP, arm FATP, leg FATP, older age, trunk FATP, and female gender were directly associated with hypertension. Ensemble methods such as voting and stacking had the best performance for hypertension prediction. Stacking showed an accuracy rate of 79%. Conclusion: By using machine learning methods, we found that BIA-derived body composition indices predict hypertension with an acceptable accuracy. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Hypertension is one of the most important and preventable causes of cardiovascular disease (CVD), stroke, chronic kidney disease, and dementia which caused approximately 8.5 million deaths in 2015, in low & middle-income countries ( 1 ). Reports show that hypertension prevalence is 25% in Iran ( 2 ). Hypertension depends on well-known risk factors such as age, gender, family history, smoking, alcohol consumption, central obesity, overweight and physical inactivity ( 3 , 4 ); obesity has gained significant attention over the past years ( 5 ). Body mass index (BMI) is widely used for anthropometric measurements, and regardless of inaccuracy, it is still commonly used to determine obesity and assess health risks such as hypertension ( 6 ). Complementary measures such as waist circumference, waist-to-hip ratio (WHR), and body composition analysis improve the prognostic efficiency of BMI ( 7 ). Evidence shows that body fat distribution is a more vital determinant of cardiovascular morbidity and mortality than increased fat mass ( 8 – 10 ); further indicating that detailed assessment of body composition is beneficial for health risk estimations. In the past few years, growing number of researchers have used machine learning and data mining algorithms to diagnose and treat health conditions such as heart ( 11 ) and brain ( 12 ) diseases. Their non-invasive nature and accuracy have enabled health professionals to quickly identify at-risk individuals and use more efficient preventive and managing strategies ( 13 ). In this study, we used machine learning approaches to investigate whether BIA-derived body composition indices predict hypertension in a cohort of patients. Methods Study design and participants Fasa cohort study ( 14 ) recruited at least 10,000 people and assessed predisposing factors for non-communicable diseases in rural regions of Fasa, Iran. In the present study, we used a subset of their data of 4663 records in which 2156 were male, 1099 had HTN, and the age range was 35–70. hypertension diagnosis was based on the blood pressure threshold defined by ACC/AHA guidelines ( 15 ). All participants had given informed consent, and the Shiraz University of Medical Sciences ethics committee approved this study. Body composition analysis Body composition analysis was performed using eight electrodes (Tanita Segmental Body Composition Analyzer BC-418 MA Tanita Corp, Japan) BIA machines. The following variables were measured: 1. Fat mass (FATM): Total fat mass (FATM), Left and Right Leg Fat Mass (LLFATM & RLFATM), Left and Right Arm Fat Mass (LAFATM & RAFATM), and Trunk Fat Mass (TRFATM) 2. Fat percentage (FATP): Total Fat Percentage (TFATP), Left and Right Leg Fat Percentage (LLFATP & RLFATP), Left and Right Arm Fat Percentage (LAFATP & RAFATP), Trunk Fat Percentage (TRFATP) Fat percentage is calculated as (Fat Mass)/Weight × 100 3. Fat-free mass (FFM): Total Fat-Free Mass (FFM), Left and Right Leg Fat-Free Mass (LLFFM & RLFFM), Left and Right Arm Fat-Free Mass (LAFFM & RAFFM), Trunk Fat-free Mass (TRFFM) 4. Basal metabolic rate (BMR) Dataset Our dataset included 4663 records; 2156 were male, and 1099 were hypertensive. Input features were: age (Between 35 and 70), gender ID (1: male, 2: female), BMR, FATM, FATP, FFM, LLFATP, RLFATP, LLFFM, RLFFM, LLFATM, RLFATM, LAFATP, RAFATP, LAFATM, RAFATM, LAFFM, RAFFM, TRFATP, TRFATM, and TRFFM. It is note that Institutional approval was granted for the use of the patient datasets in research studies for diagnostic and therapeutic purposes. Approval was granted on the grounds of existing datasets. Informed consent was obtained from all of the patients in this study. All methods were carried out in accordance with relevant guidelines and regulations. Ethical approval for the use of these data was obtained from the Tehran Omid hospital. Investigated machine learning and data mining algorithms We selected the algorithms used in this research from some of the most efficient classification algorithms, such as Support Vector Classifier (SVC) ( 16 ), Decision Tree (DT) ( 17 ), Stochastic Gradient Descend (SGD) Classifier ( 18 ), Logistic Regression (LR) ( 19 ), Gaussian Naïve Bayes (GNB) ( 20 ), K-Nearest Neighbor (K-NN) ( 21 ), Multi-Layer Perceptron (MLP) ( 22 ), Random Forest (RF) ( 23 ), Gradient Boosting (GB) ( 24 ), Histogram-based Gradient Boosting (HGB) ( 25 ), Bagging ( 26 ), Extra Tree (ET) ( 27 ), Ada Boost ( 28 ), Voting ( 29 ) and Stacking ( 30 ). These algorithms are briefly explained, and the references required for comprehensive learning about them are introduced. In the following part, we introduce metrics for evaluating the effectiveness of algorithms. To classify the data, SVC tries to find the best hyperplane to separate the different classes. The criterion to evaluate the hyper-plane is maximizing the distance between it and the sample points. SVC has a limitation compensated by the Support Vector Machine (SVM) non-linearly. It is the difference between SVC and SVM. In SVC, the hyper-plane classifies the data linearly. However, in SVM, the algorithm separates the dataset non-linearly ( 31 ). DT belongs to supervised learning algorithms used for classification and regression. This learning method tries to create a model that can predict the value of a target feature by learning some decision rules inferred from the features of samples ( 32 ). SGD Classifier is a linear classifier optimized by the SGD ( 33 ). LR is a classification algorithm used in machine learning; it uses a logistic function to model the dependent variable. This variable can only have two values. So, LR is only used in solving problems with binary target features. Also, the sigmoid function in LR maps the predicted values to the probabilities ( 34 ). GNB is a probabilistic classification algorithm that performs the classification of samples using the Bayes theorem. It assumes that the variables are independent of each other. This algorithm requires training data to estimate the parameters needed for classification. Since its implementation is simple, it is used to solve many classification problems ( 20 ). K-NN algorithm is a non-parametric, supervised classifier that uses proximity to perform classification. In this algorithm, the assumption is that similar points are located near each other. A class label is assigned to a sample based on the majority vote between K nearer samples around it ( 35 ). MLP is a supervised learning algorithm that tries to learn a function based on a data set. Then, it uses this function to predict the class of a new sample. This algorithm has a network structure consisting of several layers of nodes. Each layer is connected to the next layer in the network. Nodes in the first layer represent input data. Other nodes map inputs to outputs by linearly combining them using a set of weights and a bias and applying an activation function ( 36 ). RF is an ensemble learning method for classification. This algorithm works based on a structure that consists of many decision trees. It is created based on training data. The output of this algorithm is the class that most trees suggest. This algorithm is suitable when the training set is over-fitting. Random forest performance is usually better than decision tree classifiers, but this performance improvement usually depends on the data type ( 37 ). Another machine learning algorithm is GB . This algorithm provides a prediction model in the form of a set of weak prediction models, usually decision trees. It is one of the most popular methods of structured classification and predictive regression modeling and can cover a wide range of data sets. However, this method is too slow to train, mainly when used on large data sets (number of samples > = 10000). In order to solve this problem, the trees added to the set are trained by discretization (binning) of continuous input variables to several hundred unique values ( 24 ). This modification dramatically increases the speed of the algorithm execution compared to the Gradient Boosting Classifier. Gradient boosting ensembles that implement this technique are referred to as HGB sets. It also can manage missing values. During training, at each split point, the tree learns whether samples with missing values should go to the left or right child based on the potential gain. Consequently, samples with missing values can be assigned to the left or right child. If there are no missing values for a given feature during training, samples with missing values are mapped to whichever child has the most samples ( 25 ). A bagging classifier is an ensemble meta-classifier that consists of a set of base classifiers applied to random subsets of the original dataset. Then, each of these classifiers' results is collected, and a final prediction is derived according to them. Each base classifier is trained in parallel with a training set. The training set for each base classifier is independent of each other. Much of the original data may be repeated in the resulting training set, while other data may be omitted ( 38 ). ET classifier is an ensemble learning technique, also known as an extremely randomized tree classifier. This algorithm uses the results of several uncorrelated decision trees collected in a forest to perform the classification process. The performance of this algorithm is very similar to an RF classification. However, building decision trees in the forest is different from RF. Each decision tree is built from the original training sample in this algorithm. Then, at each test node, each tree is presented with a random sample of some features from the feature set. Each decision tree must select the best feature for splitting the data based on mathematical criteria such as the Gini index. Random selection of samples leads to multiple uncorrelated decision trees ( 27 ). An Adaptive Boosting or Adaboost (for short) classifier is a meta-classifier algorithm. This ensemble algorithm starts by fitting a classifier on the original data set. It then tries to classify the same data set again using additional copies of the classifier, except that the weights of the misclassified samples are adjusted so that subsequent classifiers focus more on complex cases. The output of these classifiers is combined in a weighted summation and creates the final output of the main classifier ( 39 ). The voting classifier is a meta-classifier that trains base models. Then, according to the results of the based models, it guesses the final result. Aggregation of the results of base learners is done in two ways: hard voting and soft voting. In the former, voting is done based on the output class declared by each base learner, while in the latter, the output class is based on the probability predicted by the base classes ( 40 ). Stacking or Stacked Generalization is an ensemble meta-learning algorithm. Using this algorithm makes it possible to learn how to combine the results of two or more basic machine learning algorithms in the best possible way. The advantage of this method is that it can use the capabilities of a wide range of well-performing algorithms and make a prediction that is better than the performance of basic algorithms ( 41 ). We will apply these algorithms to our dataset. However, we will use some preprocesses before applying these algorithms. Data preprocessing A first, the correlation matrix of the patient features is estimated and shown in Fig. 1 . Based on this matrix, evidently some features have a highly positive correlation. Consequently, several features can be removed to make the classification process more accurate. For example, in leg analyses, both mass (RLFATM) and percentage (RLFATP) are measured but there is a high positive dependency between these features in the correlation matrix. As a result, only the percentage was used according to physicians' opinions. Also, the values have been measured for both the right and left sides of the body (e.g. RLFATP and LLFATP), but they also show a high positive dependency. So, we only selected the measured values of the right side of the body, and removed its prefix (R). Putting all the above together, the final list of features used in this research is FFM, FATP, BMR, arm fat-free mass (AFFM), TRFFM, leg fat-free mass (LFFM), Male, Female, Age, TRFATP, leg fat percentage (LFATP), and arm fat percentage (AFATP). The correlation between these features and the target feature is shown in Fig. 2 . Evaluation Metrics In this research, we used the confusion matrix to test and compare the algorithms' effectiveness. This matrix is a popular metric to evaluate the performance of binary and multi-class classification problems. Figure 3 shows a confusion matrix. The confusion matrix shows how many outputs are correctly classified and how many are misclassified. In this table, "TN", for true negative, shows how many negative samples are correctly classified. Similarly, "TP" stands for true positive and indicates how many positive samples are correctly classified. The term "FP" stands for false positive and represents the number of samples misclassified as positive. Finally, "FN" stands for false negative and indicates the number of positive samples misclassified as negative. Based on the values of this matrix, one of the most common metrics used for evaluating classification algorithms –accuracy- is calculated based on Eq. ( 1 ) (42). $$accuracy=\frac{TP+TN}{TP+TN+FP+FN}$$ 1 Precision, sensitivity (or recall), specificity, and F1-score are some other performance metrics that are very popular. They are calculated according to the following equations: \(Macro Average Precision=\frac{\frac{TP}{TP+FP}+\frac{TN}{TN+FN}}{2}\) (2) \(Macro Average Sensitivity = \frac{\frac{TP}{TP+FN}+\frac{TN}{TN+FP}}{2}\) (3) \(Specificity=\frac{TN}{TN+FP}\) (4) \(F1-score=2*\frac{precision*sensitivity }{precision+sensitivity}\) (5) Finally, the classification algorithms mentioned above will be compared according to the flowchart shown in Fig. 4 and using these metrics. Experimental Results In this section we report and compare the results of applying classification algorithms mentioned in the methodology section. These algorithms are implemented in Python version 3.10,0 and run in Windows 11 operating system. The default settings of the algorithms are used in this research, except those listed in Table 1 . Table 1 Settings of the used algorithms Algorithm Settings SVC Kernel: linear, random state: 0 DT Criterion: entropy, maximum depth: 3 SGD Estimator: SVM LR Maximum iteration: 500 GNB --- K-NN-3 Number of neighbors: 3 K-NN-4 Number of neighbors: 4 K-NN-5 Number of neighbors: 5 MLP Maximum iteration: 1000, random state: 0 RF --- GB --- Bagging Base estimator: Decision Tree ET Maximum depth: 4, number of estimators: 20, random state: 0 Ada Boost Number of estimators: 100 HGB Maximum iteration: 200 Voting 1 Estimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3), Extra Trees (maximum depth: 4, number of estimators: 20, random state:0)}, Voting: soft Voting 2 Estimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3)}, Voting: soft Stacking 1 Estimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3), Extra Trees (maximum depth: 4, number of estimators: 20, random state: 0)} Stacking 2 Estimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3)} Table 2 presents the data of confusion matrices. Table 3 lists the accuracy, precision, recall, f1-score, and AUC of these algorithms, ordered by accuracy. Table 2 Confusion matrix data of classification algorithms Algorithm TN TP FP FN SVC 713 18 7 195 DT 708 16 12 197 SGD 703 16 17 197 LR 715 5 5 208 GNB 654 50 66 163 K-NN-3 629 50 91 163 K-NN-4 691 24 29 189 K-NN-5 659 37 61 176 MLP 720 0 0 213 RF 673 41 47 172 GB 689 37 31 176 Bagging 663 47 57 166 ET 720 2 0 211 Ada Boost 695 20 25 193 HGB 665 42 55 171 Voting 1 709 16 11 197 Voting 2 709 19 11 194 Stacking 1 705 28 15 185 Stacking 2 704 26 16 187 Table 3 performance metrics of different classification algorithms ordered by accuracy ascendingly Algorithm Accuracy Macro Average Precision Macro Average Sensitivity Specificity F1-score AUC K-NN-3 0.73 0.57 0.55 0.87 0.28 0.57 K-NN-5 0.74 0.58 0.54 0.92 0.23 0.58 GNB 0.75 0.62 0.57 0.91 0.30 0.63 Bagging 0.76 0.63 0.57 0.92 0.30 0.63 HGB 0.76 0.61 0.56 0.92 0.27 0.65 SGD 0.77 0.63 0.53 0.98 0.14 0.61 LR 0.77 0.64 0.51 0.99 0.04 0.64 K-NN-4 0.77 0.62 0.54 0.96 0.18 0.57 MLP 0.77 0.39 0.50 1.00 --- 0.60 RF 0.77 0.63 0.56 0.93 0.27 0.64 ET 0.77 0.89 0.50 1.00 0.02 0.69 Adaboost 0.77 0.61 0.53 0.97 0.15 0.64 SVC 0.78 0.75 0.53 0.99 0.14 0.64 DT 0.78 0.68 0.53 0.98 0.14 0.67 GB 0.78 0.67 0.57 0.96 0.26 0.67 Voting 1 0.78 0.69 0.53 0.98 0.14 0.69 Voting 2 0.78 0.71 0.54 0.98 0.16 0.68 Stacking 2 0.78 0.70 0.55 0.98 0.2 0.68 Stacking 1 0.79 0.72 0.56 0.98 0.22 0.69 Stacking 1 has the best accuracy rate. Stacking 2, Voting 2, Voting 1, GB, DT, and SVC are in the following ranks. Stacking 1 also has good performance according to other metrics parameters. Thus, it is clear that the performance of ensemble learning algorithms is better than others. Discussion In the present study and a cohort population, we used machine learning methods and found that BIA-derived body composition indices predict hypertension with an acceptable accuracy. FATP, AFFM, BMR, FFM, TRFFM, AFATP, LFATP, and older age were the top features in hypertension prediction. FATP, AFATP, LFATP, TRFATP, higher age, and female gender directly associated with HTN. But, FFM, AFFM, LFFM, TRFFM, BMR, and male gender were inversely linked to HTN. Ensemble machine learning methods such as voting and stacking had the best performance for predicting hypertension, and the latter had an accuracy rate of 79%. Total FATP and FFM Various other studies confirm the direct link of body fat mass (and percentage) with blood pressure ( 43 – 45 ). Park et al. ( 46 ), in a prospective cohort study, showed that a high body fat percentage (more than 19.9% in men and 32.5% in women) was associated with an increased risk of incident hypertension regardless of BMI, waist circumference, and WHR. Although body fat mass and percentage are superior to BMI in morbidities risk assessment, a study ( 47 ) on Iranian population showed that BMI predicts CVD better than body fat percentage. Another study ( 48 ) on American postmenopausal women with normal BMI found no relation between whole-body fat mass and percentage of CVD risk; although regional body fat had significant associations. These discrepancies may be due to different analysis methods of body composition, and ethnicity. Contrary to our results, some investigations in adult and pediatric populations established that FFM is positively related to systolic, diastolic, or mean blood pressure ( 49 – 55 ). Korhonen et al. ( 50 ) attribute this finding to muscle mass properties; during daytime and contraction, skeletal muscles release myokines that may increase blood pressure. This explanation confirms the findings of Ye et al. ( 44 ) in a Chinese population: total skeletal mass (TSM) indices -primarily arm lean body mass- are positively associated with blood pressure, pre-HTN, and HTN. Trunk FATP and FFM Previous studies have established the positive association of TRFATM with hypertension and CVD ( 56 ), and our data further support that BIA-measured abdominal adiposity is positively associated with hypertension ( 57 ). Chen et al. ( 48 ) assessed CVD incidence in postmenopausal women with normal BMI during a median of 17.9 years. The authors used Dual X-ray Absorptiometry (DXA) and found that higher TRFATP and lower LFATP were associated with higher CVD risk. In an opinion survey ( 55 ), using DXA body measurement and machine learning methods, researchers depicted that TRFAT correlates with both mean systolic and diastolic pressure -the same as our findings. The authors have not provided trunk lean body mass results but declare that total lean body mass positively correlates with mean systolic blood pressure. In general, evidence is lacking about the association between TRFFM and hypertension risk. Appendicular FATP and FFM There are conflicting data about arm and leg fat association with HTN. In a study of 3130 Chinese participants by Ye et al. ( 44 ), fat mass percentage and lean body mass, especially in the arm, were positively associated with increased blood pressure. Nevertheless, leg lean mass showed no significant association with systolic and diastolic pressure. In another study ( 58 ) on 399 participants, authors showed that: 1) arm fat was a positive predictor for blood pressure, 2) after full adjustment, loss of lean leg mass directly correlated with reductions in systolic blood pressure, 3) loss of leg fat and lean mass had direct beneficial changes in markers of CVD risk. More conflicting results exist: positive association of mid-upper arm circumference with increased hypertension risk ( 59 ), and significant inverse association between the leg and arm total fat percentage with hypertension ( 60 ). The exact mechanism by which LFATP and LFFM modulate blood pressure is still unclear. Regional fat deposition in the legs, mainly subcutaneous, reduces fatty acid turnover and downregulates triglyceride production in the blood. So, it acts as a “metabolic sink” and preserves other tissues from lipotoxicity, protects endothelium against damage, and maintains elasticity and compliance of arterioles ( 58 , 61 ). Another possible mechanism is that as subcutaneous fat, it may decrease the activation of renin-angiotensin-aldosterone and the sympathetic system ( 61 ). Also, for FFM, some studies declare that muscle mass has a protective role in blood pressure ( 62 , 63 ). But, Ye et al. ( 44 ) suggest that previous studies on appendicular lean mass or skeletal muscle did not control fat mass and fat distribution in their analysis, leading to inaccurate results. Gender and age Sex differences did not predict hypertension in our study population; however, the association was negative in males and positive in females. Previous studies showed that in men, lower body fat (thigh or gynoid) had a more protective effect on cardio-metabolic risks, such as elevated blood pressure; the effects of sex hormones on subcutaneous fat mass in these regions might explain this sex difference ( 64 ). Based on our results, age had a positive association with hypertension. Likewise, a study on the Chinese population age indicated an independent association in both men and women with hypertension ( 65 ). However, results are not always positive; in a study performed on Brazilian children and adolescents, regardless of sex, the authors observed no significant association between age and systolic blood pressure ( 66 ). BMR Our study demonstrated a strong inverse relationship between BMR and hypertension, but this is not reported elsewhere. A study in Bangladeshi adults showed a positive relation between BMR and blood pressure, suggesting that upregulated BMR may elevate blood pressure by accelerating thyroid hormone levels and increasing sympathetic tone and oxidative damage ( 67 ). Further investigation is required. Strengths and limitations The implication of machine learning in a cohort of patients is the main strength of our study. Machine learning methods are more precise than traditional ones, so we believe that our findings can resolve the conflicting results regarding our research question. Nevertheless, this study has some limitations including lack of data about the use of anti-hypertensive drugs and other anthropometric indices such as waist circumference. Also, BIA of TRFAT do not differentiate between visceral and subcutaneous abdominal adipose tissues. However, we aimed to use an available method for body composition analysis and BIA is a simple, safe, and readily available method –unlike DEXA, CT scan, and MRI. We suggest that future prospective studies use machine learning methods and body composition analyses to predict hypertension among different ethnic groups. Conclusion Given that body fat and its distribution are risk factors for hypertension, we used machine learning methods to study these relations. With an acceptable accuracy, we confirmed that BIA-derived body composition predicts hypertension. Also, total and regional FATP, higher age, and female gender had a positive relation with hypertension while it was the exact contrary for total and regional FFM, BMR, and male gender. Declarations Availability of data Data are available from the authors upon reasonable request from the corresponding author, Hamed Bazrafshan Drissi. Competing interests The authors declare that they have no competing interests. Funding None Author contributions M.A.N., S.J., A.A., M.S., A.D., M.M., M.R., G.G., R.A., M.B., and H.B. are helping with writing the text. H.B.D. and S.M.S.I. have supervised us during the writing of the paper with their invaluable comments. At the end, this manuscript has resulted by the collaboration of all authors. References Zhou B, Perel P, Mensah GA, Ezzati M. Global epidemiology, health burden and effective interventions for elevated blood pressure and hypertension. Nat Rev Cardiol. 2021;18(11):785–802. Oori MJ, Mohammadi F, Norozi K, Fallahi-Khoshknab M, Ebadi A, Gheshlagh RG. Prevalence of HTN in Iran: Meta-analysis of Published Studies in 2004–2018. Curr Hypertens Rev. 2019;15(2):113–22. Qiu L, Wang W, Sa R, Liu F. Prevalence and Risk Factors of Hypertension, Diabetes, and Dyslipidemia among Adults in Northwest China. Int J Hypertens. 2021;2021:5528007. Yagoub U, Saiyed NS, Al Qahtani B, Al Zahrani AM, Birema Y, Al Hariri I. Investigating the incidence and risk factors of hypertension: A multicentre retrospective cohort study in Tabuk, Saudi Arabia. PLOS ONE. 2022;17(1):e0262259. Goto K, Yokokawa H, Fukuda H, Saita M, Hamada C, Hisaoka T, et al. An association between subcutaneous fat mass accumulation and hypertension. J Gen Fam Med. 2021;22(4):209–17. Nuttall FQ. Body Mass Index: Obesity, BMI, and Health: A Critical Review. Nutr Today. 2015;50(3):117–28. González-Muniesa P, Mártinez-González MA, Hu FB, Després JP, Matsuzawa Y, Loos RJF, et al. Obesity. Nat Rev Dis Primers. 2017;3:17034. Blüher M, Laufs U. New concepts for body shape-related cardiovascular risk: role of fat distribution and adipose tissue function. Eur Heart J. 2019;40(34):2856–8. Yano Y, Vongpatanasin W, Ayers C, Turer A, Chandra A, Carnethon MR, et al. Regional Fat Distribution and Blood Pressure Level and Variability: The Dallas Heart Study. Hypertension. 2016;68(3):576–83. Gowri SM, Antonisamy B, Geethanjali FS, Thomas N, Jebasingh F, Paul TV, et al. Distinct opposing associations of upper and lower body fat depots with metabolic and cardiovascular disease risk markers. Int J Obes (Lond). 2021;45(11):2490–8. Joloudari JH, Mojrian S, Nodehi I, Mashmool A, Zadegan ZK, Shirkharkolaie SK, et al. Application of Artificial Intelligence Techniques for Automated Detection of Myocardial Infarction: a Review. 2022. Shoeibi A, Ghassemi N, Khodatars M, Moridian P, Alizadehsani R, Zare A, et al. Detection of epileptic seizures on EEG signals using ANFIS classifier, autoencoders and fuzzy entropies. Biomedical Signal Processing and Control. 2022;73:103417. Chowdhury MZI, Naeem I, Quan H, Leung AA, Sikdar KC, O'Beirne M, et al. Prediction of hypertension using traditional regression and machine learning models: A systematic review and meta-analysis. PLoS One. 2022;17(4):e0266334. Farjam M, Bahrami H, Bahramali E, Jamshidi J, Askari A, Zakeri H, et al. A cohort study protocol to analyze the predisposing factors to common chronic non-communicable diseases in rural areas: Fasa Cohort Study. BMC Public Health. 2016;16(1):1090. Whelton PK, Carey RM, Aronow WS, Casey DE, Jr., Collins KJ, Dennison Himmelfarb C, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Hypertension. 2018;71(6):e13-e115. Lau K, Wu Q. Online training of support vector classifier. Pattern Recognition. 2003;36(8):1913–20. Safavian SR, Landgrebe D. A survey of decision tree classifier methodology. IEEE transactions on systems, man, and cybernetics. 1991;21(3):660–74. Song S, Chaudhuri K, Sarwate AD, editors. Stochastic gradient descent with differentially private updates. 2013 IEEE global conference on signal and information processing; 2013: IEEE. Hosmer Jr DW, Lemeshow S, Sturdivant RX. Applied logistic regression: John Wiley & Sons; 2013. Ontivero-Ortega M, Lage-Castellanos A, Valente G, Goebel R, Valdes-Sosa M. Fast Gaussian Naïve Bayes for searchlight classification analysis. Neuroimage. 2017;163:471–9. Wu Y, Ianakiev K, Govindaraju V. Improved k-nearest neighbor classification. Pattern recognition. 2002;35(10):2311–8. Taud H, Mas J. Multilayer perceptron (MLP). Geomatic approaches for modeling land change scenarios: Springer; 2018. p. 451–5. Biau G, Scornet E. A random forest guided tour. Test. 2016;25(2):197–227. Bentéjac C, Csörgő A, Martínez-Muñoz G. A comparative analysis of gradient boosting algorithms. Artificial Intelligence Review. 2021;54(3):1937–67. Friedman JH. Stochastic gradient boosting. Computational statistics & data analysis. 2002;38(4):367–78. Zareapoor M, Shamsolmoali P. Application of credit card fraud detection: Based on bagging ensemble classifier. Procedia computer science. 2015;48(2015):679–85. Abhishek L, editor Optical character recognition using ensemble of SVM, MLP and extra trees classifier. 2020 International Conference for Emerging Technology (INCET); 2020: IEEE. Schapire RE. Explaining adaboost. Empirical inference: Springer; 2013. p. 37–52. Parhami B. Voting algorithms. IEEE transactions on reliability. 1994;43(4):617–29. Sikora R. A modified stacking ensemble machine learning algorithm using genetic algorithms. Handbook of research on organizational transformations through big data analytics: IGi Global; 2015. p. 43–53. Alizadehsani R, Zangooei MH, Hosseini MJ, Habibi J, Khosravi A, Roshanzamir M, et al. Coronary artery disease detection using computational intelligence methods. Knowledge-Based Systems. 2016;109:187–97. Alizadehsani R, Abdar M, Roshanzamir M, Khosravi A, Kebria PM, Khozeimeh F, et al. Machine learning-based coronary artery disease diagnosis: A comprehensive review. Computers in biology and medicine. 2019;111:103346. Kabir F, Siddique S, Kotwal MRA, Huda MN, editors. Bangla text document categorization using stochastic gradient descent (sgd) classifier. 2015 International Conference on Cognitive Computing and Information Processing (CCIP); 2015: IEEE. Ayoobi N, Sharifrazi D, Alizadehsani R, Shoeibi A, Gorriz JM, Moosaei H, et al. Time series forecasting of new cases and new deaths rate for COVID-19 using deep learning methods. Results in Physics. 2021;27:104495. Alizadehsani R, Khosravi A, Roshanzamir M, Abdar M, Sarrafzadegan N, Shafie D, et al. Coronary artery disease detection using artificial intelligence techniques: A survey of trends, geographical differences and diagnostic features 1991–2020. Computers in Biology and Medicine. 2021;128:104095. Shoeibi A, Ghassemi N, Khodatars M, Jafari M, Moridian P, Alizadehsani R, et al. Applications of epileptic seizures detection in neuroimaging modalities using deep learning techniques: methods, challenges, and future works. arXiv preprint arXiv:210514278. 2021. Khozeimeh F, Sharifrazi D, Izadi NH, Joloudari JH, Shoeibi A, Alizadehsani R, et al. Combining a convolutional neural network with autoencoders to predict the survival chance of COVID-19 patients. Scientific Reports. 2021;11(1):1–18. Nahavandi D, Alizadehsani R, Khosravi A, Acharya UR. Application of artificial intelligence in wearable devices: Opportunities and challenges. Computer Methods and Programs in Biomedicine. 2022;213:106541. Asgharnezhad H, Shamsi A, Alizadehsani R, Khosravi A, Nahavandi S, Sani ZA, et al. Objective evaluation of deep uncertainty predictions for covid-19 detection. Scientific Reports. 2022;12(1):1–11. Moridian P, Ghassemi N, Jafari M, Salloum-Asfar S, Sadeghi D, Khodatars M, et al. Automatic Autism Spectrum Disorder Detection Using Artificial Intelligence Methods with MRI Neuroimaging: A Review. arXiv preprint arXiv:220611233. 2022. Khozeimeh F, Sharifrazi D, Izadi NH, Joloudari JH, Shoeibi A, Alizadehsani R, et al. RF-CNN-F: random forest with convolutional neural network features for coronary artery disease diagnosis based on cardiac magnetic resonance. Scientific Reports. 2022;12(1):1–12. Kakhi K, Alizadehsani R, Kabir HD, Khosravi A, Nahavandi S, Acharya UR. The internet of medical things and artificial intelligence: trends, challenges, and opportunities. Biocybernetics and Biomedical Engineering. 2022. Li R, Tian Z, Wang Y, Liu X, Tu R, Wang Y, et al. The Association of Body Fat Percentage With Hypertension in a Chinese Rural Population: The Henan Rural Cohort Study. Frontiers in Public Health. 2020;8. Ye S, Zhu C, Wei C, Yang M, Zheng W, Gan D, et al. Associations of Body Composition with Blood Pressure and Hypertension. Obesity. 2018;26(10):1644–50. Chen M, Liu J, Ma Y, Li Y, Gao D, Chen L, et al. Association between Body Fat and Elevated Blood Pressure among Children and Adolescents Aged 7–17 Years: Using Dual-Energy X-ray Absorptiometry (DEXA) and Bioelectrical Impedance Analysis (BIA) from a Cross-Sectional Study in China. International Journal of Environmental Research and Public Health. 2021;18(17):9254. Park SK, Ryoo JH, Oh CM, Choi JM, Chung PW, Jung JY. Body fat percentage, obesity, and their relation to the incidental risk of hypertension. J Clin Hypertens (Greenwich). 2019;21(10):1496–504. Sheibani H, Esmaeili H, Tayefi M, Saberi-Karimian M, Darroudi S, Mouhebati M, et al. A comparison of body mass index and percent body fat as predictors of cardiovascular risk factors. Diabetes Metab Syndr. 2019;13(1):570–5. Chen GC, Arthur R, Iyengar NM, Kamensky V, Xue X, Wassertheil-Smoller S, et al. association between regional body fat and cardiovascular disease risk among postmenopausal women with normal body mass index. Eur Heart J. 2019;40(34):2849–55. He H, Pan L, Du J, Jin Y, Wang L, Jia P, et al. Effect of fat mass index, fat free mass index and body mass index on childhood blood pressure: a cross-sectional study in south China. Transl Pediatr. 2021;10(3):541–51. Korhonen PE, Mikkola T, Kautiainen H, Eriksson JG. Both lean and fat body mass associate with blood pressure. Eur J Intern Med. 2021;91:40–4. Rao KM, Arlappa N, Radhika MS, BalaKrishna N, Laxmaiah A, Brahmam GNV. Correlation of Fat Mass Index and Fat-Free Mass Index with percentage body fat and their association with hypertension among urban South Indian adult men and women. Annals of Human Biology. 2012;39(1):54–8. Takase M, Nakamura T, Tsuchiya N, Kogure M, Itabashi F, Narita A, et al. association between the combined fat mass and fat-free mass index and hypertension: The Tohoku Medical Megabank Community-based Cohort Study. Clin Exp Hypertens. 2021;43(7):610–21. Vaziri Y, Bulduk S, Shadman Z, Bulduk EO, Hedayati M, Koc H, et al. Lean Body Mass as a Predictive Value of Hypertension in Young Adults, in Ankara, Turkey. Iran J Public Health. 2015;44(12):1643–54. Xu R, Zhang X, Zhou Y, Wan Y, Gao X. Percentage of free fat mass is associated with elevated blood pressure in healthy Chinese children. Hypertension Research. 2019;42(1):95–104. Nath T, Ahima RS, Santhanam P. DXA measured body composition predicts blood pressure using machine learning methods. J Clin Hypertens (Greenwich). 2020;22(6):1098–100. Goswami B, Reang T, Sarkar S, Sengupta S, Bhattacharjee B. Role of body visceral fat in hypertension and dyslipidemia among the diabetic and nondiabetic ethnic population of Tripura-A comparative study. J Family Med Prim Care. 2020;9(6):2885–90. Takeoka A, Tayama J, Yamasaki H, Kobayashi M, Ogawa S, Saigo T, et al. Intra-abdominal fat accumulation is a hypertension risk factor in young adulthood: A cross-sectional study. Medicine (Baltimore). 2016;95(45):e5361. Clifton PM. Relationship Between Changes in Fat and Lean Depots Following Weight Loss and Changes in Cardiovascular Disease Risk Markers. J Am Heart Assoc. 2018;7(8). Hou Y, Jia X, Xuan L, Zhu W, Deng C, Wang L, et al. Association between mid-upper arm circumference and cardiometabolic risk in Chinese population: a cross-sectional study. BMJ Open. 2019;9(9):e028904. Visaria A, Lo D, Maniar P, Dave B, Joshi P. Leg and arm adiposity is inversely associated with diastolic hypertension in young and middle-aged United States adults. Clin Hypertens. 2022;28(1):3. Porter SA, Massaro JM, Hoffmann U, Vasan RS, O'Donnel CJ, Fox CS. Abdominal subcutaneous adipose tissue: a protective fat depot? Diabetes Care. 2009;32(6):1068–75. AlKaabi LA, Ahmed LS, Al Attiyah MF, Abdel-Rahman ME. Predicting hypertension using machine learning: Findings from Qatar Biobank Study. PLOS ONE. 2020;15(10):e0240370. Butcher JT, Mintz JD, Larion S, Qiu S, Ruan L, Fulton DJ, et al. Increased Muscle Mass Protects Against Hypertension and Renal Injury in Obesity. J Am Heart Assoc. 2018;7(16):e009358. Yang Y, Xie M, Yuan S, Zeng Y, Dong Y, Wang Z, et al. Sex differences in the associations between adiposity distribution and cardiometabolic risk factors in overweight or obese individuals: a cross-sectional study. BMC Public Health. 2021;21(1):1232. Liu Y, Li Y, He J, Ma P, Yu L, Zheng Q, et al. Gender Stratified Analyses of the Association of Skinfold Thickness with Hypertension: A Cross-Sectional Study in General Northeastern Chinese Residents. Int J Environ Res Public Health. 2018;15(12). Zaniqueli D, Alvim RO, Baldo MP, Morra EA, Mill JG. Muscle mass is the main somatic growth indicator associated with increasing blood pressure with age in children and adolescents. J Clin Hypertens (Greenwich). 2020;22(10):1908–14. Ali N, Mahmood S, Manirujjaman M, Perveen R, Al Nahid A, Ahmed S, et al. Hypertension prevalence and influence of basal metabolic rate on blood pressure among adult students in Bangladesh. BMC Public Health. 2017;18(1):58. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 27 Apr, 2023 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Major revision 24 Feb, 2023 Reviews received at journal 09 Feb, 2023 Reviewers agreed at journal 07 Feb, 2023 Reviewers invited by journal 07 Feb, 2023 Editor assigned by journal 07 Feb, 2023 Editor invited by journal 07 Nov, 2022 Submission checks completed at journal 07 Nov, 2022 First submitted to journal 03 Nov, 2022 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-2232998","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":150011417,"identity":"312e8c48-bdf5-453c-81cb-fe37cebcda26","order_by":0,"name":"Mohammad Ali Nematollahi","email":"","orcid":"","institution":"Fasa University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"Ali","lastName":"Nematollahi","suffix":""},{"id":150011419,"identity":"80434407-a580-4c75-ad49-8c014ec452f6","order_by":1,"name":"Soodeh Jahangiri","email":"","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Soodeh","middleName":"","lastName":"Jahangiri","suffix":""},{"id":150011420,"identity":"af9fe2ff-7f29-4582-a2a0-35ccaf66ebea","order_by":2,"name":"Arefeh Asadollahi","email":"","orcid":"","institution":"Fasa University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Arefeh","middleName":"","lastName":"Asadollahi","suffix":""},{"id":150011421,"identity":"641239bf-33a9-4513-b8d5-f6e50ac2e924","order_by":3,"name":"Maryam Salimi","email":"","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Maryam","middleName":"","lastName":"Salimi","suffix":""},{"id":150011422,"identity":"c1bdaf47-1dfb-4635-97d7-573b4bc2a5c7","order_by":4,"name":"Azizallah Dehghan","email":"","orcid":"","institution":"Fasa University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Azizallah","middleName":"","lastName":"Dehghan","suffix":""},{"id":150011426,"identity":"d348663a-a0eb-4f80-bcf7-96d3e5164c7f","order_by":5,"name":"Mina Mashayekh","email":"","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Mina","middleName":"","lastName":"Mashayekh","suffix":""},{"id":150011427,"identity":"ec33af62-85d7-424b-b619-3247dafc744a","order_by":6,"name":"Mohamad Roshanzamir","email":"","orcid":"","institution":"Fasa University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Mohamad","middleName":"","lastName":"Roshanzamir","suffix":""},{"id":150011428,"identity":"ad951355-8393-4532-9538-d08789e2d66e","order_by":7,"name":"Ghazal Gholamabbas","email":"","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Ghazal","middleName":"","lastName":"Gholamabbas","suffix":""},{"id":150011429,"identity":"3229aab2-169e-4c65-b85f-847128df7680","order_by":8,"name":"Roohallah Alizadehsani","email":"","orcid":"","institution":"Deakin University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Roohallah","middleName":"","lastName":"Alizadehsani","suffix":""},{"id":150011430,"identity":"3b2896cc-572a-4372-9285-a531bc816e7d","order_by":9,"name":"Mehdi Bazrafshan","email":"","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Mehdi","middleName":"","lastName":"Bazrafshan","suffix":""},{"id":150011431,"identity":"b9db5595-a059-49a2-b3b9-782fc3a509c3","order_by":10,"name":"Hanieh Bazrafshan","email":"","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Hanieh","middleName":"","lastName":"Bazrafshan","suffix":""},{"id":150011432,"identity":"7fb9e024-47ce-4878-b872-93a0e48a81f3","order_by":11,"name":"Hamed Bazrafshan Drissi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABCElEQVRIiWNgGAWjYBACAxDB2wAiGR8wA0k5BgYeorUwG4C0GJOuJbGBkBZz9tOJD97usLM3bz/M/Lmg5k76huNnDz74wGAnp9uAXYtlT+5mw7lnkpllziSzSc849ix3w5m8ZMMZDMnGZgdwOOxA7jZp3jZmNgmG/GPMPGyHczccyDGT5mE4kLgNl5bzb7f/5m2r55Hgf8z8meff4XSD828IaLmRu42Zt+2whIREMgPQusMJBjcI2XLj7WbJuWeOG0hIPGaT5u07bDjzxhtjwxkGePxyPnfjh7c7qu0l+JOBDvt2WJ7vfI7hgw8VdnK4tGACBbBKA2KVg4B8AymqR8EoGAWjYCQAABsjX2FI17RoAAAAAElFTkSuQmCC","orcid":"","institution":"Shiraz University of Medical Sciences","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Hamed","middleName":"Bazrafshan","lastName":"Drissi","suffix":""},{"id":150011433,"identity":"ad05bc4f-a3a4-4d5e-ae22-1dc1d0d65aa2","order_by":12,"name":"Sheikh Mohammed Shariful Islam","email":"","orcid":"","institution":"Deakin University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Sheikh","middleName":"Mohammed Shariful","lastName":"Islam","suffix":""}],"badges":[],"createdAt":"2022-11-03 07:59:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-2232998/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-2232998/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-023-34127-6","type":"published","date":"2023-04-27T20:35:47+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":28876511,"identity":"2ae763e4-dfae-48f8-bcc1-5cb641269bf0","added_by":"auto","created_at":"2022-11-09 21:24:22","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":367058,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003ecorrelation matrix of the investigated features\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-2232998/v1/328b7ff6bcd71b1d702bf59d.png"},{"id":28877141,"identity":"5eebdcb7-ad07-46df-aaa7-6c5e1c8f942d","added_by":"auto","created_at":"2022-11-09 21:32:22","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":24182,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe correlation between selected features and the target feature in this research\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-2232998/v1/8b80ec3473ad16717b816e33.png"},{"id":28876510,"identity":"c913d5cc-af3e-4546-9173-ed5ff189a035","added_by":"auto","created_at":"2022-11-09 21:24:22","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":11110,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eConfusion matrix and its data\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-2232998/v1/28906d895ae03e5c08c306af.png"},{"id":28877140,"identity":"36fc1ee8-d3c9-4bdb-b4d8-ec69605d63ba","added_by":"auto","created_at":"2022-11-09 21:32:22","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":21875,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe flowchart of the methodology used in this research\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-2232998/v1/fc2a881108e38629ccb45fb3.png"},{"id":44726888,"identity":"2e119647-84cf-4e4a-b208-3720ca177eb6","added_by":"auto","created_at":"2023-10-16 20:50:02","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":810522,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-2232998/v1/645af1a1-215c-413f-b3f1-65e0e142a5ae.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Body composition predicts hypertension using machine learning methods: A Cohort Study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHypertension is one of the most important and preventable causes of cardiovascular disease (CVD), stroke, chronic kidney disease, and dementia which caused approximately 8.5\u0026nbsp;million deaths in 2015, in low \u0026amp; middle-income countries (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Reports show that hypertension prevalence is 25% in Iran (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Hypertension depends on well-known risk factors such as age, gender, family history, smoking, alcohol consumption, central obesity, overweight and physical inactivity (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e); obesity has gained significant attention over the past years (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBody mass index (BMI) is widely used for anthropometric measurements, and regardless of inaccuracy, it is still commonly used to determine obesity and assess health risks such as hypertension (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Complementary measures such as waist circumference, waist-to-hip ratio (WHR), and body composition analysis improve the prognostic efficiency of BMI (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). Evidence shows that body fat distribution is a more vital determinant of cardiovascular morbidity and mortality than increased fat mass (\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e); further indicating that detailed assessment of body composition is beneficial for health risk estimations.\u003c/p\u003e \u003cp\u003eIn the past few years, growing number of researchers have used machine learning and data mining algorithms to diagnose and treat health conditions such as heart (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e) and brain (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e) diseases. Their non-invasive nature and accuracy have enabled health professionals to quickly identify at-risk individuals and use more efficient preventive and managing strategies (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn this study, we used machine learning approaches to investigate whether BIA-derived body composition indices predict hypertension in a cohort of patients.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv class=\"Section2\" id=\"Sec3\"\u003e\n \u003ch2\u003eStudy design and participants\u003c/h2\u003e\n \u003cp\u003eFasa cohort study (\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e) recruited at least 10,000 people and assessed predisposing factors for non-communicable diseases in rural regions of Fasa, Iran. In the present study, we used a subset of their data of 4663 records in which 2156 were male, 1099 had HTN, and the age range was 35\u0026ndash;70. hypertension diagnosis was based on the blood pressure threshold defined by ACC/AHA guidelines (\u003cspan class=\"CitationRef\"\u003e15\u003c/span\u003e). All participants had given informed consent, and the Shiraz University of Medical Sciences ethics committee approved this study.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv class=\"Section2\" id=\"Sec4\"\u003e\n \u003ch2\u003eBody composition analysis\u003c/h2\u003e\n \u003cp\u003eBody composition analysis was performed using eight electrodes (Tanita Segmental Body Composition Analyzer BC-418 MA Tanita Corp, Japan) BIA machines. The following variables were measured:\u003c/p\u003e\u003cspan\u003e\n \u003cp\u003e1. Fat mass (FATM): Total fat mass (FATM), Left and Right Leg Fat Mass (LLFATM \u0026amp; RLFATM), Left and Right Arm Fat Mass (LAFATM \u0026amp; RAFATM), and Trunk Fat Mass (TRFATM)\u003c/p\u003e\n \u003c/span\u003e \u003cspan\u003e\n \u003cp\u003e2. Fat percentage (FATP): Total Fat Percentage (TFATP), Left and Right Leg Fat Percentage (LLFATP \u0026amp; RLFATP), Left and Right Arm Fat Percentage (LAFATP \u0026amp; RAFATP), Trunk Fat Percentage (TRFATP)\u003c/p\u003e\n \u003c/span\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eFat percentage is calculated as (Fat Mass)/Weight \u0026times; 100\u003c/p\u003e\n \u003c/div\u003e\u003cspan\u003e\n \u003cp\u003e3. Fat-free mass (FFM): Total Fat-Free Mass (FFM), Left and Right Leg Fat-Free Mass (LLFFM \u0026amp; RLFFM), Left and Right Arm Fat-Free Mass (LAFFM \u0026amp; RAFFM), Trunk Fat-free Mass (TRFFM)\u003c/p\u003e\n \u003c/span\u003e \u003cspan\u003e\n \u003cp\u003e4. Basal metabolic rate (BMR)\u003c/p\u003e\n \u003c/span\u003e\n\u003c/div\u003e\n\u003cdiv class=\"Section2\" id=\"Sec5\"\u003e\n \u003ch2\u003eDataset\u003c/h2\u003e\n \u003cp\u003eOur dataset included 4663 records; 2156 were male, and 1099 were hypertensive. Input features were: age (Between 35 and 70), gender ID (1: male, 2: female), BMR, FATM, FATP, FFM, LLFATP, RLFATP, LLFFM, RLFFM, LLFATM, RLFATM, LAFATP, RAFATP, LAFATM, RAFATM, LAFFM, RAFFM, TRFATP, TRFATM, and TRFFM.\u003c/p\u003e\n \u003cp\u003eIt is note that Institutional approval was granted for the use of the patient datasets in research studies for diagnostic and therapeutic purposes. Approval was granted on the grounds of existing datasets. Informed consent was obtained from all of the patients in this study. All methods were carried out in accordance with relevant guidelines and regulations. Ethical approval for the use of these data was obtained from the Tehran Omid hospital.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv class=\"Section2\" id=\"Sec6\"\u003e\n \u003ch2\u003eInvestigated machine learning and data mining algorithms\u003c/h2\u003e\n \u003cp\u003eWe selected the algorithms used in this research from some of the most efficient classification algorithms, such as Support Vector Classifier (SVC) (\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e), Decision Tree (DT) (\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e), Stochastic Gradient Descend (SGD) Classifier (\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e), Logistic Regression (LR) (\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e), Gaussian Na\u0026iuml;ve Bayes (GNB) (\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e), K-Nearest Neighbor (K-NN) (\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e), Multi-Layer Perceptron (MLP) (\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e), Random Forest (RF) (\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e), Gradient Boosting (GB) (\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e), Histogram-based Gradient Boosting (HGB) (\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e), Bagging (\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e), Extra Tree (ET) (\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e), Ada Boost (\u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e), Voting (\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e) and Stacking (\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eThese algorithms are briefly explained, and the references required for comprehensive learning about them are introduced. In the following part, we introduce metrics for evaluating the effectiveness of algorithms.\u003c/p\u003e\n \u003cp\u003eTo classify the data, \u003cstrong\u003eSVC\u003c/strong\u003e tries to find the best hyperplane to separate the different classes. The criterion to evaluate the hyper-plane is maximizing the distance between it and the sample points. SVC has a limitation compensated by the Support Vector Machine (SVM) non-linearly. It is the difference between SVC and SVM. In SVC, the hyper-plane classifies the data linearly. However, in SVM, the algorithm separates the dataset non-linearly (\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eDT\u003c/strong\u003e belongs to supervised learning algorithms used for classification and regression. This learning method tries to create a model that can predict the value of a target feature by learning some decision rules inferred from the features of samples (\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eSGD\u003c/strong\u003e Classifier is a linear classifier optimized by the SGD (\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eLR\u003c/strong\u003e is a classification algorithm used in machine learning; it uses a logistic function to model the dependent variable. This variable can only have two values. So, LR is only used in solving problems with binary target features. Also, the sigmoid function in LR maps the predicted values to the probabilities (\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eGNB\u003c/strong\u003e is a probabilistic classification algorithm that performs the classification of samples using the Bayes theorem. It assumes that the variables are independent of each other. This algorithm requires training data to estimate the parameters needed for classification. Since its implementation is simple, it is used to solve many classification problems (\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eK-NN\u003c/strong\u003e algorithm is a non-parametric, supervised classifier that uses proximity to perform classification. In this algorithm, the assumption is that similar points are located near each other. A class label is assigned to a sample based on the majority vote between K nearer samples around it (\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eMLP\u003c/strong\u003e is a supervised learning algorithm that tries to learn a function based on a data set. Then, it uses this function to predict the class of a new sample. This algorithm has a network structure consisting of several layers of nodes. Each layer is connected to the next layer in the network. Nodes in the first layer represent input data. Other nodes map inputs to outputs by linearly combining them using a set of weights and a bias and applying an activation function (\u003cspan class=\"CitationRef\"\u003e36\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eRF\u003c/strong\u003e is an ensemble learning method for classification. This algorithm works based on a structure that consists of many decision trees. It is created based on training data. The output of this algorithm is the class that most trees suggest. This algorithm is suitable when the training set is over-fitting. Random forest performance is usually better than decision tree classifiers, but this performance improvement usually depends on the data type (\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eAnother machine learning algorithm is \u003cstrong\u003eGB\u003c/strong\u003e. This algorithm provides a prediction model in the form of a set of weak prediction models, usually decision trees. It is one of the most popular methods of structured classification and predictive regression modeling and can cover a wide range of data sets. However, this method is too slow to train, mainly when used on large data sets (number of samples\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;10000). In order to solve this problem, the trees added to the set are trained by discretization (binning) of continuous input variables to several hundred unique values (\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e). This modification dramatically increases the speed of the algorithm execution compared to the Gradient Boosting Classifier. Gradient boosting ensembles that implement this technique are referred to as \u003cstrong\u003eHGB\u003c/strong\u003e sets. It also can manage missing values. During training, at each split point, the tree learns whether samples with missing values should go to the left or right child based on the potential gain. Consequently, samples with missing values can be assigned to the left or right child. If there are no missing values for a given feature during training, samples with missing values are mapped to whichever child has the most samples (\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eA \u003cstrong\u003ebagging\u003c/strong\u003e classifier is an ensemble meta-classifier that consists of a set of base classifiers applied to random subsets of the original dataset. Then, each of these classifiers\u0026apos; results is collected, and a final prediction is derived according to them. Each base classifier is trained in parallel with a training set. The training set for each base classifier is independent of each other. Much of the original data may be repeated in the resulting training set, while other data may be omitted (\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eET\u003c/strong\u003e classifier is an ensemble learning technique, also known as an extremely randomized tree classifier. This algorithm uses the results of several uncorrelated decision trees collected in a forest to perform the classification process. The performance of this algorithm is very similar to an RF classification. However, building decision trees in the forest is different from RF. Each decision tree is built from the original training sample in this algorithm. Then, at each test node, each tree is presented with a random sample of some features from the feature set. Each decision tree must select the best feature for splitting the data based on mathematical criteria such as the Gini index. Random selection of samples leads to multiple uncorrelated decision trees (\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eAn Adaptive Boosting or \u003cstrong\u003eAdaboost\u003c/strong\u003e (for short) classifier is a meta-classifier algorithm. This ensemble algorithm starts by fitting a classifier on the original data set. It then tries to classify the same data set again using additional copies of the classifier, except that the weights of the misclassified samples are adjusted so that subsequent classifiers focus more on complex cases. The output of these classifiers is combined in a weighted summation and creates the final output of the main classifier (\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eThe voting\u003c/strong\u003e classifier is a meta-classifier that trains base models. Then, according to the results of the based models, it guesses the final result. Aggregation of the results of base learners is done in two ways: hard voting and soft voting. In the former, voting is done based on the output class declared by each base learner, while in the latter, the output class is based on the probability predicted by the base classes (\u003cspan class=\"CitationRef\"\u003e40\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eStacking\u003c/strong\u003e or Stacked Generalization is an ensemble meta-learning algorithm. Using this algorithm makes it possible to learn how to combine the results of two or more basic machine learning algorithms in the best possible way. The advantage of this method is that it can use the capabilities of a wide range of well-performing algorithms and make a prediction that is better than the performance of basic algorithms (\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eWe will apply these algorithms to our dataset. However, we will use some preprocesses before applying these algorithms.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv class=\"Section2\" id=\"Sec7\"\u003e\n \u003ch2\u003eData preprocessing\u003c/h2\u003e\n \u003cp\u003eA first, the correlation matrix of the patient features is estimated and shown in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eBased on this matrix, evidently some features have a highly positive correlation. Consequently, several features can be removed to make the classification process more accurate. For example, in leg analyses, both mass (RLFATM) and percentage (RLFATP) are measured but there is a high positive dependency between these features in the correlation matrix. As a result, only the percentage was used according to physicians\u0026apos; opinions. Also, the values have been measured for both the right and left sides of the body (e.g. RLFATP and LLFATP), but they also show a high positive dependency. So, we only selected the measured values of the right side of the body, and removed its prefix (R). Putting all the above together, the final list of features used in this research is FFM, FATP, BMR, arm fat-free mass (AFFM), TRFFM, leg fat-free mass (LFFM), Male, Female, Age, TRFATP, leg fat percentage (LFATP), and arm fat percentage (AFATP). The correlation between these features and the target feature is shown in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv class=\"Section3\" id=\"Sec8\"\u003e\n \u003ch2\u003eEvaluation Metrics\u003c/h2\u003e\n \u003cp\u003eIn this research, we used the confusion matrix to test and compare the algorithms\u0026apos; effectiveness. This matrix is a popular metric to evaluate the performance of binary and multi-class classification problems. Figure 3 shows a confusion matrix.\u003c/p\u003e\n \u003cp\u003eThe confusion matrix shows how many outputs are correctly classified and how many are misclassified. In this table, \u0026quot;TN\u0026quot;, for true negative, shows how many negative samples are correctly classified. Similarly, \u0026quot;TP\u0026quot; stands for true positive and indicates how many positive samples are correctly classified. The term \u0026quot;FP\u0026quot; stands for false positive and represents the number of samples misclassified as positive. Finally, \u0026quot;FN\u0026quot; stands for false negative and indicates the number of positive samples misclassified as negative. Based on the values of this matrix, one of the most common metrics used for evaluating classification algorithms \u0026ndash;accuracy- is calculated based on Eq. (\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e) (42).\u003c/p\u003e\n \u003cdiv class=\"Equation\" id=\"Equ1\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e$$accuracy=\\frac{TP+TN}{TP+TN+FP+FN}$$\u003c/div\u003e\n \u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\n \u003c/div\u003e\n \u003cp\u003ePrecision, sensitivity (or recall), specificity, and F1-score are some other performance metrics that are very popular. They are calculated according to the following equations:\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable border=\"1\" id=\"Tabb\"\u003e\n \u003ccolgroup cols=\"2\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(Macro Average Precision=\\frac{\\frac{TP}{TP+FP}+\\frac{TN}{TN+FN}}{2}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e(2)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(Macro Average Sensitivity = \\frac{\\frac{TP}{TP+FN}+\\frac{TN}{TN+FP}}{2}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(Specificity=\\frac{TN}{TN+FP}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(4)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(F1-score=2*\\frac{precision*sensitivity }{precision+sensitivity}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e(5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eFinally, the classification algorithms mentioned above will be compared according to the flowchart shown in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e and using these metrics.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eExperimental Results\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eIn this section we report and compare the results of applying classification algorithms mentioned in the methodology section. These algorithms are implemented in Python version 3.10,0 and run in Windows 11 operating system. The default settings of the algorithms are used in this research, except those listed in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable border=\"1\" id=\"Tab1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSettings of the used algorithms\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"2\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAlgorithm\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSettings\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eKernel: linear, random state: 0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCriterion: entropy, maximum depth: 3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSGD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEstimator: SVM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMaximum iteration: 500\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e---\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNumber of neighbors: 3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNumber of neighbors: 4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNumber of neighbors: 5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMLP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMaximum iteration: 1000, random state: 0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e---\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e---\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBagging\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBase estimator: Decision Tree\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eET\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMaximum depth: 4, number of estimators: 20, random state: 0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAda Boost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNumber of estimators: 100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHGB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMaximum iteration: 200\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoting 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEstimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3), Extra Trees (maximum depth: 4, number of estimators: 20, random state:0)}, Voting: soft\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoting 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEstimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3)}, Voting: soft\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStacking 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEstimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3), Extra Trees (maximum depth: 4, number of estimators: 20, random state: 0)}\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStacking 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEstimators: {SVC (kernel: linear, probability: True, random state: 0), Gradient Boosting, Decision Tree (criterion: entropy, maximum depth: 3)}\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eTable \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e presents the data of confusion matrices. Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e lists the accuracy, precision, recall, f1-score, and AUC of these algorithms, ordered by accuracy.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable border=\"1\" id=\"Tab2\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eConfusion matrix data of classification algorithms\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"5\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAlgorithm\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTN\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTP\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFP\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFN\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e713\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e195\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e708\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e197\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSGD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e703\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e197\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e715\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e208\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e654\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e163\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e629\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e163\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e691\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e189\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e659\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e176\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMLP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e720\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e213\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e673\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e172\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e689\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e176\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBagging\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e663\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e166\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eET\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e720\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e211\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAda Boost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e695\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e193\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHGB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e665\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e171\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoting 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e709\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e197\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoting 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e709\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e194\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStacking 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e705\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e185\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStacking 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e704\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e187\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"char\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable border=\"1\" id=\"Tab3\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eperformance metrics of different classification algorithms ordered by accuracy ascendingly\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"7\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAlgorithm\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMacro Average Precision\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMacro Average Sensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eF1-score\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGNB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBagging\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHGB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.65\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSGD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.04\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eK-NN-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMLP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e---\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eET\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAdaboost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.61\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSVC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoting 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVoting 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eStacking 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.68\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eStacking 1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.79\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.72\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.56\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.98\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.22\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.69\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eStacking 1 has the best accuracy rate. Stacking 2, Voting 2, Voting 1, GB, DT, and SVC are in the following ranks. Stacking 1 also has good performance according to other metrics parameters. Thus, it is clear that the performance of ensemble learning algorithms is better than others.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn the present study and a cohort population, we used machine learning methods and found that BIA-derived body composition indices predict hypertension with an acceptable accuracy. FATP, AFFM, BMR, FFM, TRFFM, AFATP, LFATP, and older age were the top features in hypertension prediction. FATP, AFATP, LFATP, TRFATP, higher age, and female gender directly associated with HTN. But, FFM, AFFM, LFFM, TRFFM, BMR, and male gender were inversely linked to HTN. Ensemble machine learning methods such as voting and stacking had the best performance for predicting hypertension, and the latter had an accuracy rate of 79%.\u003c/p\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eTotal FATP and FFM\u003c/h2\u003e \u003cp\u003eVarious other studies confirm the direct link of body fat mass (and percentage) with blood pressure (\u003cspan additionalcitationids=\"CR44\" citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e). Park et al. (\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e), in a prospective cohort study, showed that a high body fat percentage (more than 19.9% in men and 32.5% in women) was associated with an increased risk of incident hypertension regardless of BMI, waist circumference, and WHR. Although body fat mass and percentage are superior to BMI in morbidities risk assessment, a study (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e) on Iranian population showed that BMI predicts CVD better than body fat percentage. Another study (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e) on American postmenopausal women with normal BMI found no relation between whole-body fat mass and percentage of CVD risk; although regional body fat had significant associations. These discrepancies may be due to different analysis methods of body composition, and ethnicity.\u003c/p\u003e \u003cp\u003eContrary to our results, some investigations in adult and pediatric populations established that FFM is positively related to systolic, diastolic, or mean blood pressure (\u003cspan additionalcitationids=\"CR50 CR51 CR52 CR53 CR54\" citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e). Korhonen et al. (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e) attribute this finding to muscle mass properties; during daytime and contraction, skeletal muscles release myokines that may increase blood pressure. This explanation confirms the findings of Ye et al. (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e) in a Chinese population: total skeletal mass (TSM) indices -primarily arm lean body mass- are positively associated with blood pressure, pre-HTN, and HTN.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eTrunk FATP and FFM\u003c/h2\u003e \u003cp\u003ePrevious studies have established the positive association of TRFATM with hypertension and CVD (\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e), and our data further support that BIA-measured abdominal adiposity is positively associated with hypertension (\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e). Chen et al. (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e) assessed CVD incidence in postmenopausal women with normal BMI during a median of 17.9 years. The authors used Dual X-ray Absorptiometry (DXA) and found that higher TRFATP and lower LFATP were associated with higher CVD risk.\u003c/p\u003e \u003cp\u003eIn an opinion survey (\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e), using DXA body measurement and machine learning methods, researchers depicted that TRFAT correlates with both mean systolic and diastolic pressure -the same as our findings. The authors have not provided trunk lean body mass results but declare that total lean body mass positively correlates with mean systolic blood pressure. In general, evidence is lacking about the association between TRFFM and hypertension risk.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eAppendicular FATP and FFM\u003c/h2\u003e \u003cp\u003eThere are conflicting data about arm and leg fat association with HTN. In a study of 3130 Chinese participants by Ye et al. (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e), fat mass percentage and lean body mass, especially in the arm, were positively associated with increased blood pressure. Nevertheless, leg lean mass showed no significant association with systolic and diastolic pressure. In another study (\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e) on 399 participants, authors showed that: 1) arm fat was a positive predictor for blood pressure, 2) after full adjustment, loss of lean leg mass directly correlated with reductions in systolic blood pressure, 3) loss of leg fat and lean mass had direct beneficial changes in markers of CVD risk. More conflicting results exist: positive association of mid-upper arm circumference with increased hypertension risk (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e), and significant inverse association between the leg and arm total fat percentage with hypertension (\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe exact mechanism by which LFATP and LFFM modulate blood pressure is still unclear. Regional fat deposition in the legs, mainly subcutaneous, reduces fatty acid turnover and downregulates triglyceride production in the blood. So, it acts as a \u0026ldquo;metabolic sink\u0026rdquo; and preserves other tissues from lipotoxicity, protects endothelium against damage, and maintains elasticity and compliance of arterioles (\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e). Another possible mechanism is that as subcutaneous fat, it may decrease the activation of renin-angiotensin-aldosterone and the sympathetic system (\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e). Also, for FFM, some studies declare that muscle mass has a protective role in blood pressure (\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e). But, Ye et al. (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e) suggest that previous studies on appendicular lean mass or skeletal muscle did not control fat mass and fat distribution in their analysis, leading to inaccurate results.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eGender and age\u003c/h2\u003e \u003cp\u003eSex differences did not predict hypertension in our study population; however, the association was negative in males and positive in females. Previous studies showed that in men, lower body fat (thigh or gynoid) had a more protective effect on cardio-metabolic risks, such as elevated blood pressure; the effects of sex hormones on subcutaneous fat mass in these regions might explain this sex difference (\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eBased on our results, age had a positive association with hypertension. Likewise, a study on the Chinese population age indicated an independent association in both men and women with hypertension (\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e). However, results are not always positive; in a study performed on Brazilian children and adolescents, regardless of sex, the authors observed no significant association between age and systolic blood pressure (\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003eBMR\u003c/h2\u003e \u003cp\u003eOur study demonstrated a strong inverse relationship between BMR and hypertension, but this is not reported elsewhere. A study in Bangladeshi adults showed a positive relation between BMR and blood pressure, suggesting that upregulated BMR may elevate blood pressure by accelerating thyroid hormone levels and increasing sympathetic tone and oxidative damage (\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e). Further investigation is required.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eStrengths and limitations\u003c/h2\u003e \u003cp\u003eThe implication of machine learning in a cohort of patients is the main strength of our study. Machine learning methods are more precise than traditional ones, so we believe that our findings can resolve the conflicting results regarding our research question. Nevertheless, this study has some limitations including lack of data about the use of anti-hypertensive drugs and other anthropometric indices such as waist circumference. Also, BIA of TRFAT do not differentiate between visceral and subcutaneous abdominal adipose tissues. However, we aimed to use an available method for body composition analysis and BIA is a simple, safe, and readily available method \u0026ndash;unlike DEXA, CT scan, and MRI. We suggest that future prospective studies use machine learning methods and body composition analyses to predict hypertension among different ethnic groups.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eGiven that body fat and its distribution are risk factors for hypertension, we used machine learning methods to study these relations. With an acceptable accuracy, we confirmed that BIA-derived body composition predicts hypertension. Also, total and regional FATP, higher age, and female gender had a positive relation with hypertension while it was the exact contrary for total and regional FFM, BMR, and male gender.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAvailability of data\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData are available from the authors upon reasonable request from the corresponding author, Hamed Bazrafshan Drissi.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNone\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eM.A.N., S.J., A.A., M.S., A.D., M.M., M.R., G.G., R.A., M.B., and H.B. are helping with writing the text. H.B.D. and S.M.S.I. have supervised us during the writing of the paper with their invaluable comments. At the end, this manuscript has resulted by the collaboration of all authors.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003e\u003cspan\u003eZhou B, Perel P, Mensah GA, Ezzati M. Global epidemiology, health burden and effective interventions for elevated blood pressure and hypertension. Nat Rev Cardiol. 2021;18(11):785\u0026ndash;802.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eOori MJ, Mohammadi F, Norozi K, Fallahi-Khoshknab M, Ebadi A, Gheshlagh RG. Prevalence of HTN in Iran: Meta-analysis of Published Studies in 2004\u0026ndash;2018. Curr Hypertens Rev. 2019;15(2):113\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eQiu L, Wang W, Sa R, Liu F. Prevalence and Risk Factors of Hypertension, Diabetes, and Dyslipidemia among Adults in Northwest China. Int J Hypertens. 2021;2021:5528007.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eYagoub U, Saiyed NS, Al Qahtani B, Al Zahrani AM, Birema Y, Al Hariri I. Investigating the incidence and risk factors of hypertension: A multicentre retrospective cohort study in Tabuk, Saudi Arabia. PLOS ONE. 2022;17(1):e0262259.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eGoto K, Yokokawa H, Fukuda H, Saita M, Hamada C, Hisaoka T, et al. An association between subcutaneous fat mass accumulation and hypertension. J Gen Fam Med. 2021;22(4):209\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eNuttall FQ. Body Mass Index: Obesity, BMI, and Health: A Critical Review. Nutr Today. 2015;50(3):117\u0026ndash;28.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eGonz\u0026aacute;lez-Muniesa P, M\u0026aacute;rtinez-Gonz\u0026aacute;lez MA, Hu FB, Despr\u0026eacute;s JP, Matsuzawa Y, Loos RJF, et al. Obesity. Nat Rev Dis Primers. 2017;3:17034.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eBl\u0026uuml;her M, Laufs U. New concepts for body shape-related cardiovascular risk: role of fat distribution and adipose tissue function. Eur Heart J. 2019;40(34):2856\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eYano Y, Vongpatanasin W, Ayers C, Turer A, Chandra A, Carnethon MR, et al. Regional Fat Distribution and Blood Pressure Level and Variability: The Dallas Heart Study. Hypertension. 2016;68(3):576\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eGowri SM, Antonisamy B, Geethanjali FS, Thomas N, Jebasingh F, Paul TV, et al. Distinct opposing associations of upper and lower body fat depots with metabolic and cardiovascular disease risk markers. Int J Obes (Lond). 2021;45(11):2490\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eJoloudari JH, Mojrian S, Nodehi I, Mashmool A, Zadegan ZK, Shirkharkolaie SK, et al. Application of Artificial Intelligence Techniques for Automated Detection of Myocardial Infarction: a Review. 2022.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eShoeibi A, Ghassemi N, Khodatars M, Moridian P, Alizadehsani R, Zare A, et al. Detection of epileptic seizures on EEG signals using ANFIS classifier, autoencoders and fuzzy entropies. Biomedical Signal Processing and Control. 2022;73:103417.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eChowdhury MZI, Naeem I, Quan H, Leung AA, Sikdar KC, O\u0026apos;Beirne M, et al. Prediction of hypertension using traditional regression and machine learning models: A systematic review and meta-analysis. PLoS One. 2022;17(4):e0266334.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eFarjam M, Bahrami H, Bahramali E, Jamshidi J, Askari A, Zakeri H, et al. A cohort study protocol to analyze the predisposing factors to common chronic non-communicable diseases in rural areas: Fasa Cohort Study. BMC Public Health. 2016;16(1):1090.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eWhelton PK, Carey RM, Aronow WS, Casey DE, Jr., Collins KJ, Dennison Himmelfarb C, et al. 2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA Guideline for the Prevention, Detection, Evaluation, and Management of High Blood Pressure in Adults: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines. Hypertension. 2018;71(6):e13-e115.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eLau K, Wu Q. Online training of support vector classifier. Pattern Recognition. 2003;36(8):1913\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSafavian SR, Landgrebe D. A survey of decision tree classifier methodology. IEEE transactions on systems, man, and cybernetics. 1991;21(3):660\u0026ndash;74.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSong S, Chaudhuri K, Sarwate AD, editors. Stochastic gradient descent with differentially private updates. 2013 IEEE global conference on signal and information processing; 2013: IEEE.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eHosmer Jr DW, Lemeshow S, Sturdivant RX. Applied logistic regression: John Wiley \u0026amp; Sons; 2013.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eOntivero-Ortega M, Lage-Castellanos A, Valente G, Goebel R, Valdes-Sosa M. Fast Gaussian Na\u0026iuml;ve Bayes for searchlight classification analysis. Neuroimage. 2017;163:471\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eWu Y, Ianakiev K, Govindaraju V. Improved k-nearest neighbor classification. Pattern recognition. 2002;35(10):2311\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eTaud H, Mas J. Multilayer perceptron (MLP). Geomatic approaches for modeling land change scenarios: Springer; 2018. p.\u0026nbsp;451\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eBiau G, Scornet E. A random forest guided tour. Test. 2016;25(2):197\u0026ndash;227.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eBent\u0026eacute;jac C, Cs\u0026ouml;rgő A, Mart\u0026iacute;nez-Mu\u0026ntilde;oz G. A comparative analysis of gradient boosting algorithms. Artificial Intelligence Review. 2021;54(3):1937\u0026ndash;67.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eFriedman JH. Stochastic gradient boosting. Computational statistics \u0026amp; data analysis. 2002;38(4):367\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eZareapoor M, Shamsolmoali P. Application of credit card fraud detection: Based on bagging ensemble classifier. Procedia computer science. 2015;48(2015):679\u0026ndash;85.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAbhishek L, editor Optical character recognition using ensemble of SVM, MLP and extra trees classifier. 2020 International Conference for Emerging Technology (INCET); 2020: IEEE.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSchapire RE. Explaining adaboost. Empirical inference: Springer; 2013. p.\u0026nbsp;37\u0026ndash;52.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eParhami B. Voting algorithms. IEEE transactions on reliability. 1994;43(4):617\u0026ndash;29.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSikora R. A modified stacking ensemble machine learning algorithm using genetic algorithms. Handbook of research on organizational transformations through big data analytics: IGi Global; 2015. p.\u0026nbsp;43\u0026ndash;53.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAlizadehsani R, Zangooei MH, Hosseini MJ, Habibi J, Khosravi A, Roshanzamir M, et al. Coronary artery disease detection using computational intelligence methods. Knowledge-Based Systems. 2016;109:187\u0026ndash;97.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAlizadehsani R, Abdar M, Roshanzamir M, Khosravi A, Kebria PM, Khozeimeh F, et al. Machine learning-based coronary artery disease diagnosis: A comprehensive review. Computers in biology and medicine. 2019;111:103346.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKabir F, Siddique S, Kotwal MRA, Huda MN, editors. Bangla text document categorization using stochastic gradient descent (sgd) classifier. 2015 International Conference on Cognitive Computing and Information Processing (CCIP); 2015: IEEE.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAyoobi N, Sharifrazi D, Alizadehsani R, Shoeibi A, Gorriz JM, Moosaei H, et al. Time series forecasting of new cases and new deaths rate for COVID-19 using deep learning methods. Results in Physics. 2021;27:104495.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAlizadehsani R, Khosravi A, Roshanzamir M, Abdar M, Sarrafzadegan N, Shafie D, et al. Coronary artery disease detection using artificial intelligence techniques: A survey of trends, geographical differences and diagnostic features 1991\u0026ndash;2020. Computers in Biology and Medicine. 2021;128:104095.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eShoeibi A, Ghassemi N, Khodatars M, Jafari M, Moridian P, Alizadehsani R, et al. Applications of epileptic seizures detection in neuroimaging modalities using deep learning techniques: methods, challenges, and future works. arXiv preprint arXiv:210514278. 2021.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKhozeimeh F, Sharifrazi D, Izadi NH, Joloudari JH, Shoeibi A, Alizadehsani R, et al. Combining a convolutional neural network with autoencoders to predict the survival chance of COVID-19 patients. Scientific Reports. 2021;11(1):1\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eNahavandi D, Alizadehsani R, Khosravi A, Acharya UR. Application of artificial intelligence in wearable devices: Opportunities and challenges. Computer Methods and Programs in Biomedicine. 2022;213:106541.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAsgharnezhad H, Shamsi A, Alizadehsani R, Khosravi A, Nahavandi S, Sani ZA, et al. Objective evaluation of deep uncertainty predictions for covid-19 detection. Scientific Reports. 2022;12(1):1\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eMoridian P, Ghassemi N, Jafari M, Salloum-Asfar S, Sadeghi D, Khodatars M, et al. Automatic Autism Spectrum Disorder Detection Using Artificial Intelligence Methods with MRI Neuroimaging: A Review. arXiv preprint arXiv:220611233. 2022.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKhozeimeh F, Sharifrazi D, Izadi NH, Joloudari JH, Shoeibi A, Alizadehsani R, et al. RF-CNN-F: random forest with convolutional neural network features for coronary artery disease diagnosis based on cardiac magnetic resonance. Scientific Reports. 2022;12(1):1\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKakhi K, Alizadehsani R, Kabir HD, Khosravi A, Nahavandi S, Acharya UR. The internet of medical things and artificial intelligence: trends, challenges, and opportunities. Biocybernetics and Biomedical Engineering. 2022.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eLi R, Tian Z, Wang Y, Liu X, Tu R, Wang Y, et al. The Association of Body Fat Percentage With Hypertension in a Chinese Rural Population: The Henan Rural Cohort Study. Frontiers in Public Health. 2020;8.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eYe S, Zhu C, Wei C, Yang M, Zheng W, Gan D, et al. Associations of Body Composition with Blood Pressure and Hypertension. Obesity. 2018;26(10):1644\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eChen M, Liu J, Ma Y, Li Y, Gao D, Chen L, et al. Association between Body Fat and Elevated Blood Pressure among Children and Adolescents Aged 7\u0026ndash;17 Years: Using Dual-Energy X-ray Absorptiometry (DEXA) and Bioelectrical Impedance Analysis (BIA) from a Cross-Sectional Study in China. International Journal of Environmental Research and Public Health. 2021;18(17):9254.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003ePark SK, Ryoo JH, Oh CM, Choi JM, Chung PW, Jung JY. Body fat percentage, obesity, and their relation to the incidental risk of hypertension. J Clin Hypertens (Greenwich). 2019;21(10):1496\u0026ndash;504.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSheibani H, Esmaeili H, Tayefi M, Saberi-Karimian M, Darroudi S, Mouhebati M, et al. A comparison of body mass index and percent body fat as predictors of cardiovascular risk factors. Diabetes Metab Syndr. 2019;13(1):570\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eChen GC, Arthur R, Iyengar NM, Kamensky V, Xue X, Wassertheil-Smoller S, et al. association between regional body fat and cardiovascular disease risk among postmenopausal women with normal body mass index. Eur Heart J. 2019;40(34):2849\u0026ndash;55.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eHe H, Pan L, Du J, Jin Y, Wang L, Jia P, et al. Effect of fat mass index, fat free mass index and body mass index on childhood blood pressure: a cross-sectional study in south China. Transl Pediatr. 2021;10(3):541\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKorhonen PE, Mikkola T, Kautiainen H, Eriksson JG. Both lean and fat body mass associate with blood pressure. Eur J Intern Med. 2021;91:40\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eRao KM, Arlappa N, Radhika MS, BalaKrishna N, Laxmaiah A, Brahmam GNV. Correlation of Fat Mass Index and Fat-Free Mass Index with percentage body fat and their association with hypertension among urban South Indian adult men and women. Annals of Human Biology. 2012;39(1):54\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eTakase M, Nakamura T, Tsuchiya N, Kogure M, Itabashi F, Narita A, et al. association between the combined fat mass and fat-free mass index and hypertension: The Tohoku Medical Megabank Community-based Cohort Study. Clin Exp Hypertens. 2021;43(7):610\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eVaziri Y, Bulduk S, Shadman Z, Bulduk EO, Hedayati M, Koc H, et al. Lean Body Mass as a Predictive Value of Hypertension in Young Adults, in Ankara, Turkey. Iran J Public Health. 2015;44(12):1643\u0026ndash;54.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eXu R, Zhang X, Zhou Y, Wan Y, Gao X. Percentage of free fat mass is associated with elevated blood pressure in healthy Chinese children. Hypertension Research. 2019;42(1):95\u0026ndash;104.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eNath T, Ahima RS, Santhanam P. DXA measured body composition predicts blood pressure using machine learning methods. J Clin Hypertens (Greenwich). 2020;22(6):1098\u0026ndash;100.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eGoswami B, Reang T, Sarkar S, Sengupta S, Bhattacharjee B. Role of body visceral fat in hypertension and dyslipidemia among the diabetic and nondiabetic ethnic population of Tripura-A comparative study. J Family Med Prim Care. 2020;9(6):2885\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eTakeoka A, Tayama J, Yamasaki H, Kobayashi M, Ogawa S, Saigo T, et al. Intra-abdominal fat accumulation is a hypertension risk factor in young adulthood: A cross-sectional study. Medicine (Baltimore). 2016;95(45):e5361.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eClifton PM. Relationship Between Changes in Fat and Lean Depots Following Weight Loss and Changes in Cardiovascular Disease Risk Markers. J Am Heart Assoc. 2018;7(8).\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eHou Y, Jia X, Xuan L, Zhu W, Deng C, Wang L, et al. Association between mid-upper arm circumference and cardiometabolic risk in Chinese population: a cross-sectional study. BMJ Open. 2019;9(9):e028904.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eVisaria A, Lo D, Maniar P, Dave B, Joshi P. Leg and arm adiposity is inversely associated with diastolic hypertension in young and middle-aged United States adults. Clin Hypertens. 2022;28(1):3.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003ePorter SA, Massaro JM, Hoffmann U, Vasan RS, O\u0026apos;Donnel CJ, Fox CS. Abdominal subcutaneous adipose tissue: a protective fat depot? Diabetes Care. 2009;32(6):1068\u0026ndash;75.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAlKaabi LA, Ahmed LS, Al Attiyah MF, Abdel-Rahman ME. Predicting hypertension using machine learning: Findings from Qatar Biobank Study. PLOS ONE. 2020;15(10):e0240370.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eButcher JT, Mintz JD, Larion S, Qiu S, Ruan L, Fulton DJ, et al. Increased Muscle Mass Protects Against Hypertension and Renal Injury in Obesity. J Am Heart Assoc. 2018;7(16):e009358.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eYang Y, Xie M, Yuan S, Zeng Y, Dong Y, Wang Z, et al. Sex differences in the associations between adiposity distribution and cardiometabolic risk factors in overweight or obese individuals: a cross-sectional study. BMC Public Health. 2021;21(1):1232.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eLiu Y, Li Y, He J, Ma P, Yu L, Zheng Q, et al. Gender Stratified Analyses of the Association of Skinfold Thickness with Hypertension: A Cross-Sectional Study in General Northeastern Chinese Residents. Int J Environ Res Public Health. 2018;15(12).\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eZaniqueli D, Alvim RO, Baldo MP, Morra EA, Mill JG. Muscle mass is the main somatic growth indicator associated with increasing blood pressure with age in children and adolescents. J Clin Hypertens (Greenwich). 2020;22(10):1908\u0026ndash;14.\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAli N, Mahmood S, Manirujjaman M, Perveen R, Al Nahid A, Ahmed S, et al. Hypertension prevalence and influence of basal metabolic rate on blood pressure among adult students in Bangladesh. BMC Public Health. 2017;18(1):58.\u003c/span\u003e\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-2232998/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-2232998/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eIntroduction: \u003c/strong\u003eWe used machine learning methods to investigate if body composition indices predict hypertension.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e Data from a cohort study was used, and 4663 records were included (2156 were male, 1099 with hypertension, with the age range of 35-70 years old). Body composition analysis was done using bioelectrical impedance analysis (BIA); weight, basal metabolic rate, total and regional fat percentage (FATP), and total and regional fat-free mass (FFM) were measured. We used machine learning methods such as Support Vector Classifier, Decision Tree, Stochastic Gradient Descend Classifier, Logistic Regression, Gaussian Naïve Bayes, K-Nearest Neighbor, Multi-Layer Perceptron, Random Forest, Gradient Boosting, Histogram-based Gradient Boosting, Bagging, Extra Tree, Ada Boost, Voting and Stacking to classify the investigated cases and find the most relevant features to hypertension.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eFATP, AFFM, BMR, FFM, TRFFM, AFATP, LFATP, and older age were the top features in hypertension prediction. Arm FFM, basal metabolic rate, total FFM, Trunk FFM, leg FFM, and male gender were inversely associated with hypertension, but total FATP, arm FATP, leg FATP, older age, trunk FATP, and female gender were directly associated with hypertension. Ensemble methods such as voting and stacking had the best performance for hypertension prediction. Stacking showed an accuracy rate of 79%.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion:\u003c/strong\u003e By using machine learning methods, we found that BIA-derived body composition indices predict hypertension with an acceptable accuracy.\u003c/p\u003e","manuscriptTitle":"Body composition predicts hypertension using machine learning methods: A Cohort Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2022-11-09 21:24:17","doi":"10.21203/rs.3.rs-2232998/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Major revision","date":"2023-02-24T11:58:28+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2023-02-10T02:43:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"988e85c8-c2a5-48d2-bde6-9361a8ac7d1f","date":"2023-02-07T22:35:22+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2023-02-07T19:54:15+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2023-02-07T19:53:00+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2022-11-07T07:37:29+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2022-11-07T07:33:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2022-11-03T07:52:16+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"66c5ca09-8fb7-4f63-b29a-74bf8abea188","owner":[],"postedDate":"November 9th, 2022","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":16768152,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":16768153,"name":"Health sciences/Diseases"}],"tags":[],"updatedAt":"2023-10-16T20:41:47+00:00","versionOfRecord":{"articleIdentity":"rs-2232998","link":"https://doi.org/10.1038/s41598-023-34127-6","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2023-04-27 20:35:47","publishedOnDateReadable":"April 27th, 2023"},"versionCreatedAt":"2022-11-09 21:24:17","video":"","vorDoi":"10.1038/s41598-023-34127-6","vorDoiUrl":"https://doi.org/10.1038/s41598-023-34127-6","workflowStages":[]},"version":"v1","identity":"rs-2232998","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-2232998","identity":"rs-2232998","version":["v1"]},"buildId":"7rjqhiLT3MXkJMwkYKINL","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}