Optimizing Diabetes Prediction: A Comparative Analysis of Ensemble Machine Learning Models with PSO-AdaBoost and ACO-XGBoost

Abstract

The primary objective of this study is to investigate the early identification of diabetes in order to lessen the health and economic consequences associated with this prevalent chronic condition. One notable research gap in the existing in the existing disease detection model is the restricted applicability of prediction models. The proposed diabetes prediction technique encompasses a data preprocessing procedure to assure the quality of the data. Here, comparative study has been carried out on two ensemble learning strategies, namely Particle Swarm Optimization (PSO) integrated AdaBoost, and Ant Colony Optimization (ACO) integrated XGBoost, for diabetes prediction. The optimization approaches employed in this study optimizes the model parameters in order to improve its predictive capabilities. A thorough evaluation is then undertaken by utilizing a variety of performance measures to assess its overall performance. The results indicate that the PSO-AdaBoost ensemble model outperforms the ACO-XGBoost ensemble model in terms of accuracy, precision, F1-score, and AUC score.