Missing Value Imputation on Gene Expression Data using Bee-based algorithm to Improve Classification Performance

Motivation: Existing missing value imputation methods focused on imputing the data regarding actual values. These mainly assisted towards a completion of datasets as an input for machine learning tasks. Gene expression dataset was one of expensive datasets that usually contains missing values from insufficient resolution and dust contaminants. As gene expression data were often used in machine learning for automated classification, we propose an imputation of missing value towards improvement of classification performance. The proposed method was based on Bee algorithm and K-nearest neighborhood with linear regression. Among the processes, GINI importance score was utilized in selecting values for imputation. The imputed values thus reflected on improving a discriminative power in classification tasks instead of replicating the actual values from the original dataset. Result: In this study, we evaluated the proposed method in terms of classification performance against frequently used imputation methods including K-nearest neighborhood, PCA, and MICE with 3 cancer-identifying gene expression datasets. The values in the datasets were randomly removed for 1, 2, 3, 4, 5, 10, and 20% to obtain missing value datasets in different missing amount. The results indicated that the proposed method obtained the best classification performance from all datasets comparing to other methods. In comparison to original dataset, the classification model from imputed datasets using the proposed method yielded 15-25% higher classification accuracy. The results showed that feature ranking for classification was noticeably changed as the imputed data played the role to boost a discriminative power in a classification task.