Prediction of Diabetes using Machine Learning Classifiers with Polar Bear Optimization

The COVID-19 pandemic significantly disrupted daily routines, exacerbating the risk factors associated with various diseases, including diabetes. Diabetes, recognized as a rapidly escalating global health issue, has profound implications for economic, social, and healthcare systems. Emerging studies have established a heightened risk of diabetes diagnosis among adults post-COVID-19, with individuals already diagnosed with diabetes being more susceptible to severe complications from the virus. This study focuses on feature selection utilizing Particle Swarm Optimization (PSO) and Polar Bear Optimization (PBO) algorithms, with a Logistic Regression (LR) as objective function. Four machine learning (ML) classifiers—Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF), and Light Gradient Boosting Machine (LightGBM) has been implemented here to evaluate the predictive accuracy of diabetes in terms of accuracy (Acc.), precision (Pre.), recall (Re.), F1-score (F1), brier score (BS) and area under the receiver operating characteristic curve (AUC-ROC). Notably, the LightGBM classifier optimized with PBO demonstrated superior performance, achieving classification report Acc. as 98.21%, Pre. as 98%, Re. as 98%, F1 as 98%, BS as 3.15% and AUC-ROC as 99.11%. This research aims to improve diabetes prediction accuracy, thereby enhancing early diagnosis and intervention strategies, ultimately contributing to better health outcomes and reducing the socio-economic burden of diabetes in affected communities.