A Hybrid Whale Optimization and XGBoost Framework for Accurate Prediction of Type 2 Diabetes Mellitus

Abstract

Introduaction Type 2 Diabetes Mellitus (T2DM) has become a worldwide health issue that has to be taken care of. Thus, predictive models have to be developed that are accurate and efficient to help with the early diagnosis and preventive measures. In this work, a hybrid of the Whale Optimization Algorithm (WOA) and Extreme Gradient Boosting (XGBoost) is proposed to improve T2DM prediction. Materials and Methods To optimize XGBoost hyperparameters for better generalization and fewer classification errors, the Whale Optimization Algorithm (WOA) was used. To assess the effectiveness and performance of the suggested approach, two benchmark datasets were evaluated: the PIMA Indian Diabetes dataset (PID) with 768 records and the Diabetes Risk Prediction dataset with 100,000 records. Results The WOA-XGBoost model recorded an accuracy of 98.7%, precision, recall, and F1-score of 99% for the dataset, which consists of 768 records, and an accuracy of 99.84%, precision-99.91%, recall-99.89% and F1-Score-99.9% for the dataset, which consists of 100000 records. It was observed that the proposed method performed better than the other state-of-the-art methods. Conclusion The proposed WOA-XGBoost model demonstrates highly accurate and reliable prediction performance for T2DM across both small and large datasets. These results indicate that the hybrid optimizationbased approach is practical for early diagnosis and can be valuable in real-world clinical decision-support systems.