Comparison of SVM and Random Forest with RFE for Diabetes Prediction

Abstract

Individuals with a high Body Mass Index (BMI) are among the most vulnerable groups to developing Type 2 Diabetes Mellitus due to insulin resistance caused by visceral fat accumulation. However, most existing machine learning models have been developed using general population data without considering the specific characteristics of high-risk individuals. This study aims to analyze and compare the performance of Support Vector Machine (SVM) and Random Forest (RF) algorithms in predicting diabetes risk among individuals with a BMI ≥ 25, while also evaluating the impact of Recursive Feature Elimination (RFE) on improving model performance. The Pima Indians Diabetes Dataset from the UCI Machine Learning Repository was used as the data source. After filtering records based on BMI, a total of 662 instances were included in the analysis. The preprocessing stage consisted of median imputation for invalid values, feature normalization using StandardScaler, and feature selection using RFE to select four features for each model. The dataset was divided into training and testing sets using a 70:30 ratio. Model performance was evaluated using accuracy, precision, recall, F1-score, and confusion matrix metrics. The results indicate that the RF-RFE model achieved the best performance, with an accuracy of 78%, a recall of 68% for the diabetes class, and an F1-score of 72%, representing a significant improvement over the RF model without RFE (74% accuracy and 60% recall). The combination of Random Forest and Recursive Feature Elimination proved to be the most effective approach for reducing false negatives, which is particularly important in the context of early clinical detection of diabetes