Symptom Intelligence: High-Accuracy Disease Prediction with Ensemble Learning

Accurate and timely disease prediction based on patient-reported symptoms remains a significant challenge in modern healthcare. Early diagnosis is essential for improving treatment outcomes and enhancing overall healthcare efficiency. However, many existing symptom-based diagnostic systems rely on binary symptom representation and a single machine learning model, which may limit their clinical realism, robustness, and predictive reliability. To address these limitations, this study proposes an ensemble learning framework that incorporates symptom severity for multi-disease prediction. Instead of binary encoding, symptoms are represented using weighted severity scores to better capture their clinical intensity. The proposed framework integrates three widely used classifiers Random Forest, Naïve Bayes, and Support Vector Machine and combines their predictions through a majority-voting ensemble strategy. The framework is evaluated using the publicly available Kaggle Disease Symptom Description Dataset, which contains 4,920 records across 41 diseases and 133 symptoms. Data preprocessing includes severity mapping, label encoding, and feature construction, followed by model training using an 80:20 train–test split. Experimental results show that while individual models achieve strong performance, the ensemble approach consistently provides higher accuracy, precision, recall, and F1-score. Additionally, the trained model is deployed in a web-based prototype that accepts user symptoms and generates disease predictions with recommended precautions, demonstrating the practical applicability of the proposed approach.