Smart E-Healthcare System

Computer Engineering

Healthcare systems are increasingly adopting digital solutions to assist medical professionals in early disease diagnosis and patient monitoring. Existing symptom-based diagnostic systems predominantly focus on single-disease prediction, which fails to reflect real-world clinical scenarios where patients frequently present with overlapping symptoms and comorbid conditions. This paper proposes a Smart E-Healthcare System that leverages multi-label machine learning techniques to predict multiple diseases simultaneously from patient-reported symptoms. The system employs a fuzzy symptom-matching algorithm based on Levenshtein distance similarity to handle input variations and typographical errors, combined with a SHAP-inspired explain- ability module that provides per-symptom contribution scores to enhance clinical transparency and trust. The prediction engine supports eight major diseases—Diabetes, Hypertension, Asthma, Thyroid Disorder, Anemia, Migraine, Arthritis, and Common Cold—each characterized by ten clinically relevant symptoms with associated importance weights. Model performance is eval- uated using appropriate multi-label metrics including Hamming Loss, Precision, Recall, and F1-score. The backend is imple- mented using Node.js with Express.js and a MySQL relational database, while the frontend delivers a responsive, mobile- friendly interface with integrated medicine recommendations and location-based doctor discovery. Experimental evaluation demon- strates a symptom recognition accuracy of 91%, Hamming Loss of 0.08, and macro-averaged F1-score of 0.87, with prediction response times under 50 milliseconds. The system addresses key challenges in e-health including security, scalability, and usability, making it a viable tool for preliminary multi-disease diagnostic support in real-world clinical environments.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.