MediAI: A Multimodal Artificial Intelligence System for Personalized Disease Prediction and Healthcare Assistance

Abstract

This paper presents a Personalized Medical Recommendation System (PMRS) that integrates machine learning-based disease risk prediction, explainable artificial intelligence, and personalized healthcare recommendations within a unified decision support framework. The system utilizes multi-modal data inputs, including demographic, clinical, and lifestyle attributes, to enable accurate and context-aware prediction of chronic diseases. Multiple supervised learning models, including Random Forest, Gradient Boosting, Support Vector Machine, Logistic Regression, and Decision Tree, are evaluated, with Random Forest achieving the best performance (91.2% accuracy and 91.1% F1-score).
To enhance transparency and clinical interpretability, SHAP (SHapley Additive exPlanations) is employed to provide both global and local explanations of model predictions. The system further incorporates severity risk stratification and prediction confidence estimation to support effective clinical decision-making and prioritization of high-risk patients. A personalized recommendation engine generates guideline-consistent health plans aligned with established standards such as ADA 2024 and ESC 2023. In addition, the framework integrates a geo-spatial hospital recommendation module to assist users in identifying nearby healthcare facilities based on location and service relevance.
Experimental evaluation on benchmark clinical datasets demonstrates the effectiveness, robustness, and practical applicability of the proposed system. The PMRS addresses key limitations of existing approaches by combining predictive accuracy, interpretability, and actionable healthcare guidance within a scalable and integrated architecture, thereby contributing to the development of intelligent and accessible preventive healthcare systems.