Comprehensive Data-Driven Insights into Metabolic Syndrome: Utilizing DeepGBM with Heuristic Feature Optimization
This study investigates the superior predictive performance of a DeepGBM model (combining boosting and deep learning) for identifying metabolic syndrome in the Korean adult population using KNHANES data. DeepGBM consistently showed superior performance compared to established algorithms. Feature prioritization revealed waist circumference and fasting glucose as critical predictors. This research demonstrates the potential of integrating advanced machine learning with public health data to improve early detection.