Abstract Lifestyle diseases, including type 2 diabetes, hypertension, obesity, and cardiovascular disorders, contribute significantly to global morbidity, affecting over 1 billion individuals and imposing substantial economic burdens estimated at 1. 3 trillion annually. Conventional fitness and nutrition applications rely predominantly on collaborative filtering or content-based methods, which inadequately account for user-specific constraints such as comorbidities, physical limitations, medication interactions, and dietary restrictions. This research introduces an integrated framework comprising a constraint-based recommendation engine, a conversational AI chat-box, and a smart diet scanner leveraging computer vision. The constraint engine employs knowledge-based filtering to generate tailored fitness regimens and nutritional plans compliant with clinical guidelines (e. g. , ADA for diabetes, AHA for hypertension). The AI chat-box facilitates real-time, natural language interactions for plan adjustments and motivational support, while the diet scanner enables instantaneous meal analysis via convolutional neural networks (CNNs) like YOLOv8. Simulation results on synthetic datasets (n=500 users) demonstrate a 28% increase in adherence rates and 22% improvement in simulated health outcomes (e. g. , BMI reduction) compared to baseline systems like MyFitnessPal or generic ML recommenders.
Shelke et al. (2026) studied this question.