Allergic Rhinitis Prediction Through Machine Learning with Integrating Environmental, Immunologic, and Demographic Factors

Allergic rhinitis (AR) is a widespread allergic reaction that has been shown to be impacted by the function of the immune system as well as environment and socioeconomic factors. This research is to explore the best predictive model among penalized logistic regression, random forest classifier, and XGBoost classifier, to gain insight into subjects who are susceptible to allergic rhinitis by taking advantage of the integrated data NHANES provides. The random forest model demonstrated the most stable performance. SHAP analysis provided interpretable insights at both group and individual levels, revealing that immune-related markers, including total IgE, eosinophil percentage, and the neutrophil-to-lymphocyte ratio were the strongest predictors of AR susceptibility. Environmental and socioeconomic exposures, such as cotinine levels, housing conditions, and income, also contributed substantially to the predicted risk. Overall, the findings highlighted that AR susceptibility arises from the combined influence of immunologic dysregulation and environmental stressors, underscoring the need for targeted preventive strategies.