Improving the Prediction of Diabetes in LIFECARE Cohort Using Cluster-Based Undersampling

This thesis aim is to help identify individuals that will develop diabetes (high blood glucose in the body). In order to be able to predict accurately, risk factors need to be identified and a prediction model is needed. As individuals that develop diabetes is of the minority group, there is data imbalance. Hence, to address this issue, Cluster-Based Undersampling was used. The LIFECARE Cohort is a Cohort in Sarawak which recruited healthy individuals to study the development of diesease. This thesis shares the significant factors identified using LIFECARE cohort and the step by step methodology for the proposed method ODS-CBUS.