Prediction of Diabetes Mellitus in the Upcoming Year using SMOTE and Random Forest

Diabetes Mellitus (DM) is one of the six diseases most people suffer from, which costs money for treatment. Patients with this disease increased by 2.4% of the total population in Indonesia between 2013 and 2018. This is undoubtedly very dangerous if serious precautions are not taken. DM prevention is critical to avoid increasing incidence in the upcoming years. One of the preventive measures is a prediction based on data from the results of a medical examination (medical or general checkup), which is taken for periodic examination once a year for three years (2019-2021). The data used in this study came from the Telkom Health Foundation (Yakes) Bandung. Predictions can be made using machine learning models, such as Random Forest, Naive Bayes, Decision Tree, Logistic Regression, AdaBoost, XGBoost, and SVM. This study aims to predict DM for the upcoming year with a preprocessing approach to general checkup data because the general checkup data has problems such as missing values and imbalanced data. This study uses the mean technique for handling missing values and the SMOTE technique for handling imbalanced. The prediction results show that the random forest prediction model is superior to other prediction models, with an accuracy of 98.2%, precision of 82.6%, recall of 66.4%, and f1-score of 71.4%.