SHIELD: Symptom-Based Hybrid Intelligent Early Learning for Disease Prediction

Abstract

Traditional diagnostic approaches often face delays and inaccuracies, while standalone machine learning models fail to account for individual uniqueness. The SHIELD system leverages hybrid machine-learning models to enhance disease prediction based on patient symptoms. This study integrates Gradient Boosting, Decision Trees, and Random Forest models, combining their strengths using an ensemble voting approach. A comprehensive dataset from Kaggle, enriched with symptom severity mappings, enables accurate and personalized predictions. The system delivers practical outputs, including disease names, descriptions, and home remedies, through a user-friendly web interface. Achieving an accuracy of approximately 99.59% with the ensemble model, SHIELD demonstrates its potential to revolutionize early disease detection, aligning with global health objectives.