Hierarchical Hidden Markov Model for Enhanced Diabetes Risk Stratification: Comparison with Machine learning models

Abstract

Traditional machine learning methods struggle with the dynamics of continuous glucose monitoring
(CGM) data. This study developed a Hierarchical Hidden Markov Model (HHMM) integrating CGM
features for diabetes risk profiling and compared it with standard classifiers. Data from 86 patients,
including clinical and CGM measurements, were analyzed using eight features.