Journal de santé et d'informatique médicale

A Comparative Performance Evaluation of Hybrid and Ensemble Machine Learning Models for Prediction of Asthma Morbidity

Abstract

Pooja MR and Pushpalatha MP

One of the chronic respiratory diseases that affect a large proportion of the population is Asthma. Asthma is more prevalent in children of age groups 6-14 years. Early identification of the risk factors is an important intervention towards the management of the disease as the disease is progressive in nature. In our work, we assess the performance of the two machine learning approaches with respect to their accuracy in predicting the outcome of asthma disease after identifying the critical risk factors that help in the prognosis of the disease. We perform an empirical analysis of ensemble and hybrid machine learning models to deduce the best performing approach for the prediction of the outcome of asthma. The Neyveli rural asthma dataset of India, representing cross sectional study data gathered through questionnaires formulated under ISAAC study was used to validate our approach. The outcome is predicted using both, sequential and parallel ensemble learning techniques as well as the hybrid machine learning model and we suggest the best performing ensemble learning technique on the dataset under consideration. The problem of class imbalance is well handled before presenting the data to the model as unbalanced data sets are seen to have a negative impact on classification performance.