Predictive Data Mining Approaches for Diabetes Mellitus Type II Disease

Diabetes is among the major public health problem especially in developing countries which cause by abnormal insulin secretion in human body. It is a common disease that can led to several health complications and mortality. In Malaysia, most of the cases are categorized as Diabetes Mellitus (DM) Type II. Patients with diabetes increases from year to year due to unhealthy lifestyles e.g. smoking, overweight and hypertension. Therefore, this study meant to identify the influential factors that may contribute to DM Type II by comparing the performance of different data mining approaches. Between April 2017 and November 2018, 684 patients from a public clinic participated in this retrospective cross-sectional study. Four predictive models involved in the study are Logistic Regression, Decision Tree, Naïve Bayes, and Artificial Neural Network (ANN). The error measures (Average Squared Error and Misclassification Rate) with ROC Index are used to evaluate the performance of the models. Results show that the performance of Logistic Regression-Stepwise outperformed to other predictive models with classification accurateness of 73% and able to predict positive outcome (Y=1) correctly by 90%. The significant inputs that affect DM Type II prediction (Y=1) are Hypertension and Glycated Hemoglobin (HbA1c) given the Root Mean Squared Error (RMSE) of model is 0.424. The importance of study may be able to contribute in improving the strategies and planning on diabetes diseases in Malaysia