A 2-Tier Stacking Ensemble Classifier for Disease Classification

Abstract

Diarrhea, dysentery, and dehydration are top of the list of high mortality-causing diseases in children under the age group of five. Despite the tremendous growth of machine learning in the field of medical research, still, some areas remain untouched, especially, the pediatric department. Since medical data is used in the proposed work, even a slight increase in the accuracy of the model will be of great importance. In this study, we propose a 2-tier stacking ensemble method for disease identification. Initially, the data is pre-processed and sent to train the tier-1 machine learning models. Based on the majority voting the metadata is selected and sent to the next tier and finally, at the meta-classifier level disease classification happens. Performance metrics like the accuracy, precision, recall, F1-score and mean absolute error of the individual machine learning algorithms were analyzed and were used to compare with the proposed stacking ensemble method. The results proved that the 2-tier stacking model proposed in the work shows an accuracy of 95.41%, a precision of 94.47%, a recall value of 92.78%, and an F1-score of 93.50%. The proposed model achieved a high accuracy value when compared to the other machine learning models.