Integrated Machine Learning Framework for Multi-Disease Prediction and Geolocation-Based Healthcare Recommendations

Chronic diseases such as diabetes, heart disease, breast cancer, and diabetic retinopathy continue to impose a substantial burden on healthcare systems due to delayed diagnosis and limited post-diagnostic guidance. Many existing digital health platforms focus solely on prediction while neglecting actionable follow-up, such as identifying suitable healthcare providers. This paper presents an integrated web-based machine learning framework that performs multi-disease risk prediction and provides location-based healthcare recommendations. The system supports four disease models: Logistic Regression for diabetes and diabetic retinopathy, and Random Forest classifiers for heart disease and breast cancer. Users manually input clinical parameters, which are preprocessed and evaluated using pre-trained models deployed on a centralized server. Experimental evaluation on publicly available datasets demonstrates classification accuracies of 75.32% for diabetes, 99.50% for diabetic retinopathy, 90.16% for heart disease, and 83.23% for breast cancer. Beyond prediction, the framework incorporates a geolocation module that recommends nearby hospitals and specialists based on the predicted outcome. The results indicate that combining disease prediction with post-prediction guidance improves practical usability, although clinical deployment would require validation on real-world patient data.