Naive Bayes Classification for Early Prediction of Diabetes Mellitus

Diabetes Mellitus is one of the chronic diseases with a prevalence that is increasing globally. Data from the World Health Organization (WHO) shows that the number of Diabetes Mellitus cases continues to rise each year. Early prediction of diabetes is crucial because there is an asymptomatic phase where individuals at risk do not exhibit clinical signs. Consequently, diabetes can become too severe to manage if not detected early. In this study, the Naïve Bayes classification learning algorithm is utilized. Naïve Bayes is a simple yet powerful classification method, especially for datasets with many attributes. The Naïve Bayes algorithm was tested using RapidMiner with a dataset of 235 patient blood test records and 2 testing data samples. The test results showed that the obtained accuracy reached 96.17%, indicating that this system can classify with a high level of accuracy. The system’s precision was recorded at 97.01%, indicating that most positive predictions are correct. Additionally, the recall of 96.30% demonstrates the system’s capability to detect most positive cases of diabetes mellitus. The Area Under the Curve (AUC) value of 0.985 from the ROC Curve indicates an excellent discriminative ability to distinguish between positive and negative cases.