Brain Stroke Detection and Doctor Recommendation System Using Machine Learning: An Integrated Approach to Predictive Healthcare

Abstract

Brain stroke remains one of the foremost causes of death and permanent neurological disability worldwide, imposing substantial socioeconomic burdens on healthcare systems globally. Timely and accurate prediction of stroke risk, combined with seamless referral to qualified specialists, is clinically decisive in minimizing irreversible brain damage and improving patient outcomes. This paper presents a comprehensive integrated Brain Stroke Detection and Doctor Recommendation System that unifies a machine learning–based diagnostic engine with a web-based intelligent specialist referral platform. The stroke detection module employs a Random Forest (RF) Classifier trained on the publicly available Kaggle Stroke Prediction Dataset comprising 5,110 patient records across 11 clinical and demographic features. A rigorous comparative study is conducted against Logistic Regression and Decision Tree classifiers under identical experimental conditions. The RF model achieves an overall accuracy of 95.66%, demonstrating robust performance characterized by high precision on the majority (non-stroke) class, a non-trivial challenge acknowledged by the severe class imbalance (4.9% stroke prevalence) inherent to real-world medical datasets. Upon detection of high stroke risk, the system automatically activates a doctor recommendation module that ranks neurologists and stroke specialists using a three-component hybrid algorithm combining content-based filtering, collaborative filtering, and ontology-based medical reasoning, further augmented by geolocation-based proximity scoring. The full system is implemented using Python Flask, PHP/MySQL, and a responsive HTML/CSS/JavaScript frontend, deployed locally on XAMPP for offline accessibility. Experimental results confirm that the proposed end-to-end architecture effectively bridges the critical translational gap between automated AI-based diagnosis and actionable patient care navigation, with potential for deployment in both urban hospital settings and resource-constrained rural healthcare environments.