Intelligent Heart Disease Prediction System Using AI, IoMT, and Modified Particle Swarm Optimization

Intelligent recommender systems have received substantial research attention because individuals need precise and customized recommendation features. The research creates an AI-based system that can adjust to users’ evolving habits and deal with environmental ambiguities by combining deep interaction modeling purposes, structurally balanced improvement, and multi-preference assessment. The Multi-Preference Integrated Deep Learning (MPIDL) algorithm in the first phase integrates different user preferences through attention-based networks by blending explicit and implicit feedback patterns. A framework named DLUSF operates with Deep Learning-Based User Session Framework to monitor and dynamically discover temporal changes in user preferences by joining LSTM and contextual Embeddings. The network completes consistency measures and minimizes user churn through implementation of the Dynamic Multi-Preference Balancing (DMPB) mechanism that preserves recommendation cluster harmony by managing the bias-variance trade-offs. Through its hybrid design the proposed system provides improved recommendation relevance together with diverse responses for different application areas such as e-commerce and media streaming and smart healthcare. Experiments conducted confirmed that the new system reaches better results for Accuracy, Precision, Recall and F1-Score metrics in comparison to traditional collaborative filtering and isolated deep learning systems. The system performs 96.85% better than benchmark standards through its implementation which effectively handles user session variability along with preference differences while proving the superiority of proposed methods over advanced contemporary approaches.