Boosting-Based Machine Learning Approaches for Diabetes Prediction using Indian Demographic and Health Survey Data

Traditional diagnostic techniques are time-consuming and prone to error. Performance is added however time and cost are reduced by using computer-aided diagnostic methods. This study aims to calculate how healthy boosting-based machine learning classifiers perform in automatically identifying diabetes in the Indian context. For this experiment, Indian demographic and health survey 2021 data on women aged 19 to 79 are considered. This data includes clinical, anthropometric, and biochemical components. 12,103 of the 706,504 responders who tested positive for diabetic testing were considered. The predictive exploration-based boosting machine learning models are adopted, such as adaptive boosting, categorical boosting, extreme gradient boosting, gradient boosting, and light gradient boosting models. Kernel Principal Component Analysis is employed for feature extraction. The Extreme Gradient Boosting model demonstrated strong performance on the diabetes dataset both before and after applying Kernel PCA. The metrics including accuracy, F1-score, precision, and recall for the XGBoost model showed improvements: accuracy increased from 81% to 83%, F1-score from 81% to 82%, precision from 81% to 88%, and recall from 81% to 76%.In contrast, Adaptive Boosting performed poorly, while Categorical Boosting, Gradient Boosting, and Light Gradient Boosting models achieved moderate performance levels. Overall, the model attained optimal performance across metrics such as accuracy, precision, Area Under the Curve (AUC), and recall.