Diabetes Detection Optimisation with Hyperparameter Tuning in Random Forest Algorithm

Diabetes is a common disease suffered by many people, one of which is Diabetes Mellitus. This disease is caused by disorders in the pancreas that affect the body's metabolism due to the lack of production of the hormone insulin, the use of technology that is associated with diabetes is one step to be able to classify diabetes. This study aims to develop a diabetes classification model using the Random Forest algorithm. The methods used include dataset selection from the Pima Indians Diabetes Database, data pre-processing by replacing missing values using the mean, and data balancing using the SMOTE technique. The model was then trained and evaluated using confusion matrix to measure accuracy, precision, recall, and F1-score. The results showed that the Random Forest algorithm with grid search hyperparameters produced good performance with 79% accuracy, 76% precision, 83% recall, and 80% F1-score. The conclusion of this research is that the Random Forest algorithm is effective in classifying diabetes data and shows improved performance compared to other algorithms such as Logistic Regression. This model can be used for more accurate early detection of diabetes, thus helping in early treatment and reducing the number of disabilities and deaths due to diabetes.