Smart Health Disease Prediction System: An End-to-End Ensemble Learning Approach

The escalating burden on global healthcare systems, exacerbated by a critical shortage of medical professionals, has necessitated the development of automated diagnostic tools to facilitate early disease detection. While recent literature has extensively explored Machine Learning (ML) for this purpose, existing studies often rely on single-algorithm models—such as K-Nearest Neighbors (KNN) or Support Vector Machines (SVM)—which frequently encounter performance ceilings and lack integration into actionable clinical workflows. This paper proposes a comprehensive Smart Health Disease Prediction System that bridges the gap between algorithmic prediction and practical telemedicine application. The proposed system utilizes an Ensemble Voting Classifier, integrating Random Forest, Logis- tic Regression, and SVM to mitigate individual model biases and enhance predictive robustness. Experimental results demonstrate that this ensemble approach achieves an accuracy of 94.24% and a recall of 94.88%, significantly outperforming the baseline SVM model (82.18%) and exceeding the 93.5% accuracy reported in comparable studies using Weighted KNN. Beyond prediction, the system introduces a holistic, web-based architecture devel- oped on the Flask framework. Key innovations include a Risk Stratification Module for real-time severity assessment, an NLP- driven Medical Chatbot for immediate patient guidance, and an automated Appointment Scheduling System to streamline the transition from digital triage to professional medical consultation. This end-to-end ecosystem offers a scalable solution for reducing physician workload while ensuring timely intervention for high- risk patients.