Ensemble Machine Learning Methods for Improved Diabetes Risk Assessment

Abstract

Diabetes is a pressing global health concern, as it is the leading cause of death worldwide and is closely associated with a range of severe complications, including kidney disease, vision loss, and cardiovascular disease. Early detection and effective management of this chronic condition are crucial to mitigate its adverse impacts on patient outcomes. to leverage data mining techniques for analyze a dataset of diabetes-related symptoms and demographics in order to develop a predictive model for diabetes diagnosis. The application of data mining approaches in healthcare has led to significant advancements in disease diagnosis and treatment, alleviating the burden on medical professionals. One area of growing interest is diabetes prediction, where early detection can significantly enhance treatment outcomes and management plans. This study presents a diabetes prediction model that leverages the strengths of four data mining techniques: Random Forests, AdaBoost, Gradient Boosting, and XGBoost. Our proposed model is developed and validated using a real-world dataset, demonstrating its potential to improve diabetes diagnosis and care. Furthermore, we utilize an Improved Chaotic Whale Optimization algorithm to select the most optimal features for the predictive model. The results demonstrate that the logistic regression model achieves the highest accuracy, outperforming the other data mining techniques explored. These findings underscore the potential of data mining approaches in enhancing diabetes diagnosis and management, ultimately leading to improved patient outcomes and reduced healthcare burden.