Implementation of Predictive Model for Diarrhea Among Afghanistan Children Based on Medical and Non-Medical Attributes

Childhood diarrheal illness is still prevalent in many low-income countries and Afghanistan is no exception to this. In spite of the fact that the cause of diarrhea is well understood in the field of medicine, other non-medical factors may attribute to this illness directly or indirectly. Nevertheless, studies of diarrhea in children, especially in Afghanistan, have ignored non-medical factors. The objective of this study is to implement predictive model for diarrhea in Afghanistan children where both medical and non-medical factors are considered. The dataset used is Afghanistan's Demographic and Health Survey (AfDHS) 2015. Information Gain and Correlation-based Feature Selection algorithms are utilized during the preprocessing. The five well-known machine learning algorithms, Support Vector Machine (SVM), Random Forest, Neural Network, Naïve Bayes and Decision Tree are investigated during implementation. Predictive models are evaluated by four common metrics, Accuracy, Area Under Curve (AUC), Precision, and Recall. It is found that the predictive model implemented by Random Forest yields the best overall performance. The predictive model in this work can be used as a preventive measure to detect the possibility of diarrhea in children at early stage.