A Predictive Model for Diabetes Medications Management

Abstract

The increasing usage of medical electronics in the healthcare industry leads to ease of accessing medical records. Accessing patients’ record electronically will save time for health professionals, patients and other concerned bodies. But the information that can be accessed from the electronic medical records or health management information system could not tell the patterns or future existence of particular diseases. In this study experiments were conducted to develop a predictive model for diabetes management that helps health professionals and other units to determine the patterns of the diabetes diseases. The model can predict either the diabetes medication is needed or not for a particular patient based on different features of the investigation. For development of the model different classification algorithms were enforced. From four different experiments conducted in the study, J-48 decision tree with percentage splitting approach scored the best classification accuracy. Future research can be attempted using different approaches like Artificial Neural network, Fuzzy logic and Neuro-fuzzy algorithms in order to compare the results and to enhance the accuracy of the model.