DiabetesXAI: An AI-Powered Web Framework for Predictive Diagnosis and Personalized Lifestyle Recommendations

Abstract Diabetes is a chronic condition affecting millions globally, requiring early diagnosis and effective management. This work introduces DiabetesXAI, an AI-driven web application designed for diabetes risk prediction using patient-specific data. The system features two key modules: (1) Medical Test-Based Prediction, which analyses clinical test results like insulin level, glucose, BMI, etc to predict, and (2) Lifestyle-Based Prediction, an AI chatbot that evaluates lifestyle factors such as diet, exercise, stress levels, etc to provide risk assessments and personalized recommendations. DiabetesXAI integrates machine learning models (Logistic Regression, Random Forest, XGBoost, etc.) and SHAP (Shapley Additive Explanations) to ensure transparent and interpretable predictions, aiding clinicians and users in informed decision-making. The model is trained on a diverse dataset (synthetic) covering demographics, medical history, and lifestyle factors, enhancing accuracy and reliability. Users interact through a web-based interface built with HTML, CSS, JavaScript, Django, and SQLite3, allowing seamless data input, model selection, chatbot-based evaluation, and downloadable PDF report generation with predictions and recommendations. The scalable and adaptable architecture supports multimodal data integration, improving prediction precision. By incorporating explainability, DiabetesXAI fosters trust among healthcare providers and patients, bridging a critical gap in AI-driven medical applications. This system aims to assist both medical professionals and individuals in proactive diabetes management through real-time risk assessment and preventive insights.