Personalized Disease Risk Prediction Using Secure Multi-Cloud AI Models in Healthcare

The increasing demand for privacy-preserving and accurate diagnostic systems in healthcare has motivated the development of intelligent hybrid architectures. This paper presents SPFH-AI - Secure Personalized Federated Hybrid AI, a novel framework designed for disease prediction across 41 medical conditions using 132 symptom parameters. The system combines Federated Learning, Multimodal Fusion, Personalized Attention, and Secure Communication Protocols to deliver robust classification while maintaining data confidentiality. Unlike traditional centralized models, SPFH-AI enables collaborative model training across distributed nodes without sharing raw patient data. Each module within the architecture plays a vital role-federated learning ensures privacy, personalized layers improve patient-specific learning, and fusion layers integrate diverse symptom data for enriched understanding. Comprehensive evaluations demonstrate that SPFH-AI achieves a classification accuracy of 96.9 %, surpassing existing models such as SVM (91.2%), MLP (92.7%), CNN (93.5%), BiLSTM (94.1%), Capsule Net (94.8 %), and Maxout Network (95.2 %). Performance metrics including precision (96.7 %), recall (96.6 %), and F1-score (96.6 %) affirm the model's effectiveness. ROC curve analysis and ablation studies further validate SPFH-AI's superiority in both performance and architecture design.