Application of the Naive Bayes classifier for predicting type 2 diabetes risk in outpatients at the metropolitan health center - Puno

Abstract

The performance of the Naive Bayes classifier was evaluated for predicting type 2 diabetes risk in outpatients treated at the Metropolitan Health Center of Puno. The analysis utilized a dataset of 121 patients collected in 2023, comprising clinical, anthropometric, and nutritional information. SMOTE was employed to mitigate class imbalance, and the Naive Bayes predictive probabilities were calibrated to improve reliability. In the test set, the model achieved an overall accuracy of 88%, with 100% sensitivity and 85% specificity, recording 17 true negatives, 5 true positives, 0 false negatives, and 3 false positives, demonstrating its utility as a screening tool for early identification of patients at risk for type 2 diabetes. Additionally, the results were compared with commonly used approaches in similar studies, positioning the performance of Naive Bayes within the current landscape of machine learning applications in public health. Naive Bayes represents a viable and operationally accessible alternative to support early diagnostic processes and the prioritization of preventive interventions in primary care centers.