Lifestyle-Based Diabetes Risk Progression and Early Warning System Using Machine Learning

Diabetes mellitus is a chronic metabolic disorder affecting hundreds of millions of individuals worldwide and represents a significant burden on global healthcare systems. Existing machine learning approaches largely address diabetes prediction as a binary classification task, failing to capture gradational risk and the potential impact of lifestyle modification. This paper presents a Lifestyle-Based Diabetes Risk Progression and Early Warning System that employs a Random Forest classifier trained on the Behavioral Risk Factor Surveillance System (BRFSS) 2015 dataset comprising 253,680 respondents and 21 health and lifestyle features. The proposed system advances the state of the art in three key dimensions: (1) it introduces a continuous risk scoring mechanism on a scale of 0 to 100 derived from class-conditional probabilities; (2) it incorporates multi-class classification to distinguish no diabetes, pre-diabetes, and diabetes; and (3) it provides a lifestyle-based risk progression simulation that models how incremental improvements in modifiable factors—such as body mass index, physical activity, dietary habits, and smoking status—can reduce an individual's predicted risk score over time. Feature importance analysis yields clinically interpretable explanations, identifying BMI, general health status, age, and high blood pressure as the most influential predictors. Experimental evaluation demonstrates an overall classification accuracy of 84%, supporting the system's potential as a preventive healthcare decision-support tool. Additionally, the system is extended with a hybrid data integration approach incorporating wearable sensor data such as step count, heart rate, and sleep patterns to enhance input reliability and enable real-time risk monitoring. A full-stack web application is developed to provide an interactive and accessible user interface for real-world deployment.Keywords — diabetes risk prediction, Random Forest, BRFSS dataset, lifestyle simulation, risk scoring, pre-diabetes detection, preventive healthcare, machine learning, feature importance