Elevating Disease Prediction: A Stacking Ensemble Learning Approach

Abstract

Timely and accurate disease diagnosis is critical for effective treatment and cost reduction. This project develops a web app utilizing machine learning to predict diseases from user-entered symptoms and recommend nearby specialists. An ensemble model stacking Logistic Regression, Random Forest, KNN, MLP, and SVM classifiers with XGBoost as the meta-learner was created using a diverse dataset of symptoms, diseases, and medical specialists. The Flask-based app features user authentication, symptom input, and Google Maps API integration for doctor recommendations. The ensemble model accurately predicts potential diseases and relevant specialists nearby, providing an interactive platform for instant diagnosis and treatment access to reduce complications and costs. Rigorous testing ensured seamless deployment, empowering early disease identification and access to appropriate healthcare. The ensemble achieved 97% accuracy, outperforming individual base learners, while KNN emerged as the top individual model with 92.3% accuracy.