A Comparative Analysis of Random Forest and Decision Tree Classifiers for Predicting Type 2 Diabetes using K-Fold Cross-Validation

Abstract

This study investigates the prediction accuracy of type 2 diabetes using the Pima-Indians-Diabetes dataset, employing a novel classifier that integrates Random Forest and K-fold cross-validation. Additionally, the findings are compared with predictions obtained from a Stochastic Gradient Descent (SGD) classifier. The dataset, consisting of 768 samples, was used for both training and testing. Using a significance level of α=0.05 and power=0.80 in the G*Power analysis, the results consistently approached an accuracy of 80%. The Decision Tree classifier achieved an accuracy of 86.245%, while the novel Random Forest classifier outperformed it with an accuracy of 96.179%. An Independent Samples T-Test yielded a p-value of 0.01 (p<0.05), indicating a statistically significant relationship between the performance of the Random Forest classifier, K-fold cross-validation, and the Decision Tree classifier. Overall, the Random Forest classifier demonstrated superior predictive accuracy, highlighting its effectiveness in type 2 diabetes prediction.