Penerapan Algoritma C4.5 untuk Prediksi Diabetes menggunakan Rapidminer

Abstract

Diabetes is a chronic disease whose prevalence continues to increase, so accurate and easy-to-interpret prediction methods are needed. The application of machine learning is one solution in analyzing medical data, especially for diabetes prediction. The Decision Tree algorithm was chosen because it is able to produce a transparent, easy-to-understand model, and is suitable for health data analysis that requires high interpretability. This study aims to apply the Decision Tree C4.5 algorithm to predict diabetes conditions using a diabetes dataset processed with RapidMiner software. The research stages include data preparation, determining attribute roles (set roles), dividing training and test data, building a classification model, implementing the model, and evaluating performance using the Performance module. The results showed that the Glucose attribute was the most dominant factor in the formation of the decision tree. The model produced an accuracy value of 80.00%, with very good performance in the non-diabetes class, but still has limitations in detecting all diabetes cases. In conclusion, the Decision Tree C4.5 algorithm is effective as an initial interpretive model in diabetes prediction, but still requires further development to increase the model's sensitivity.