Predictive Modelling and IoT-Based Early Intervention for Diabetes Mellitus in East and West Godavari Districts Using Clinical Big Data

The rapid growth of Diabetes Mellitus (DM) in the East and West Godavari districts of India demands advanced methods for early detection of risk. The present study has developed and validated an innovative prognostic framework which combines the static clinical data with simulated real-time activity monitoring. An attempt has been made to create a dataset by integrating The Pima Indians Diabetes Database and the Human Activity Recognition (HAR) Smartphones Dataset. A comparative analysis of machine learning classifiers shown that the Random Forest model yielded 92% accuracy with an F1-score of 0.91. The results also confirm that this data-fusion approach significantly enhances predictive power than models using only clinical data. This validated framework provides a robust, scalable tool for early risk assessment, enabling a critical shift from reactive treatment to proactive, personalized interventions and also informing targeted public health measures.