CardioExplain CDSS
A web-based explainable clinical decision support prototype designed to investigate transparent heart disease risk prediction using Gradient Boosting and SHAP.
Purpose of the Prototype
CardioExplain CDSS was developed as a research prototype to evaluate the feasibility of combining machine-learning prediction with interpretable explanations in a web environment. The system receives 13 clinical parameters, estimates the probability of heart disease, and presents the direction and magnitude of influential features.
This application is intended for methodological evaluation, software demonstration, and initial screening research. Its output must not be treated as a definitive diagnosis or used independently for clinical decisions.
Research Team
Universitas Medan Area
CV. Multimedia Teknologi Kreatif
Research collaboration, machine-learning integration, and web-based information system development.
Gradient Boosting
Ensemble of 80 decision trees with a learning rate of 0.05.
Permutation SHAP
Global and patient-specific contributions represented in the model’s log-odds space.
PHP Native + Python Flask
Research model development and web implementation with stored parameters for reproducible inference.
Research Model Summary
Duplicate records were removed before stratified data splitting to reduce the risk of information leakage. All trained tree parameters are stored with the application for reproducible PHP inference.