A modular cluster based collaborative recommender system for cardiac patients.

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Cardiovascular Disease Management

Background:

• Hospitals generate vast amounts of patient health data for clinical decisions.
• Collaborative filtering (CF) is successful in health recommender systems but faces sparsity and scalability challenges.
• Traditional CF methods can reduce accuracy and efficiency in clinical settings.

Purpose of the Study:

• To propose an improvised collaborative filtering technique using clustering and sub-clustering for health recommendations.
• To enhance the accuracy and efficiency of recommender systems for cardiovascular diseases.
• To address data sparsity and scalability issues in clinical recommender systems.

Main Methods:

• A modular clustering-based collaborative filtering approach was developed.
• The methodology involved partitioning patient data by cardiovascular disease type (angina, non-cardiac chest pain, silent ischemia, myocardial infarction).
• K-mean clustering was applied within each disease partition, followed by sub-clustering for similarity scoring.

Main Results:

• The proposed system significantly reduced the search domain and time for accurate recommendations.
• Experimental results demonstrated improved accuracy, precision, and recall values for health recommendations.
• The modular approach enhanced system scalability and efficiency.

Conclusions:

• The clustering and sub-clustering based recommender system effectively addresses limitations of traditional CF in healthcare.
• This method provides accurate and timely recommendations, crucial for cardiovascular disease management.
• The proposed system offers a scalable and efficient solution for health recommendation systems.