Empowering Diabetic Prediction through MRMR-Driven Feature Selection and Robustness of Ensemble Machine Learning

Effective diabetic prediction plays a pivotal role in healthcare, as it allows for early intervention and tailored treatment strategies, ultimately leading to enhanced patient outcomes and decreased healthcare costs. In recent times, the incorporation of machine learning models has exhibited potential in augmenting the accuracy of diabetic prediction models. This research study focuses on enhancing diabetic prediction through the synergistic application of ensemble machine learning and feature selection, particularly the Minimum Redundancy Maximum Relevance (MRMR) - driven feature selection method. By integrating MRMR - driven features into forward feature selection within ensemble machine learning models, the study aims to create a highly accurate Ensemble Machine Learning – Diabetic Prediction System (EML-DPS). The experimental results of EML-DPS demonstrates the effectiveness of the combination of ensemble machine learning models and MRMR - driven feature selection significantly improves predictive accuracy, reduces overfitting, and enhances the interpretability of the model. This approach, exemplified by Optimizable Ensemble’s remarkable ability to reach 97.3% accuracy with minimal set of essential features, achieves promising results in diabetic prediction.