Performance Evaluation of Classification Models for HIV/AIDS Dataset

Abstract

The classification of the HIV/AIDS testing result is one of the areas where machine learning can be implemented. In this work, we establish a robust methodology to categorize whether or not a person was being tested for HIV by machine learning algorithms (i.e., negative or positive). In this paper, we have used ten machine learning algorithms to classify the test results for HIV/AIDS among individuals. The classification algorithms used in this study are namely: gradient boosting (GB), random forest (RF), extra tree (ET), bagging, K-nearest neighbor (KNN), decision tree (DT), support vector machine (SVM), logistic regression (LR), AdaBoost (AB), and Naive Bayes (NB). Then, we perform classification tasks from a developing dataset for HIV/AIDS by gathering data from EDHS and validating it by domain experts. The performance of all classifiers has better results for all performance metrics. The performance of the random forest, extra tree, and decision tree model is outstanding with a training accuracy of 0.9755, and the other classifiers such as bagging, K-nearest neighbor, gradient boosting, logistic regression, SVM, AdaBoost, and Naive Bayes are scored 0.9669, 0.8822, 0.8774, 0.8285, 0.8234, 0.8086, and 0.7731, respectively.