AI-Based Disease Prediction System Using Machine Learning with Conversational Assistance and AI-Driven Recommendations

This paper presents a web-based medical diagnosisassistant that predicts probable diseases from user-reportedsymptoms using supervised machine learning. The model istrained on a large dataset containing approximately 246000 rowsfor diseases mapped across 377 symptoms. Classical machinelearning models including Na¨ıve Bayes and Random Forestwere implemented and compared. The Na¨ıve Bayes classifierachieved the best overall accuracy in our experiments. Thefinal deployed system integrates an interactive web interface,automated PDF report generation, a conversational chatbotfor natural-language symptom input, and Google Gemini-basedrecommendation generation for remedies and precautions. Wediscuss data preparation, model selection and tuning, evaluationmetrics, system deployment, ethical considerations, explainabilityvia SHAP/LIME, and future work. The system is intended asa preliminary screening tool to improve health awareness andhelp users make informed decisions before consulting healthcareprofessionals.Index Terms—Disease Prediction, Na¨ıve Bayes, Random For-est, Machine Learning, Chatbot, Gemini API, Explainable AI,Healthcare Informatics