Leveraging Multiactions to Improve Medical Personalized Ranking for Collaborative Filtering

Area of Science:

• Health Informatics
• Computer Science
• Information Retrieval

Background:

• High-quality recommendation services are crucial for web applications, including healthcare.
• Existing methods for recommendation systems include pointwise regression and pairwise ranking.
• Emerging healthcare platforms require specialized recommendation models for medical services.

Purpose of the Study:

• To develop an accurate and scalable recommendation model for medical services.
• To address the challenge of personalized healthcare service recommendations using implicit user feedback.
• To propose a novel approach that considers users' diverse actions for improved medical service ranking.

Main Methods:

• Developed Medical Bayesian Personalized Ranking (MBPR), a model integrating multiple user actions.
• MBPR is based on the principle that user preferences in one domain (e.g., healthcare) correlate with preferences in other actions.
• Utilized real-world datasets from healthcare and mobile shopping applications for evaluation.

Main Results:

• MBPR demonstrated superior accuracy compared to several state-of-the-art recommendation methods.
• Experimental results validated the model's effectiveness on diverse, real-world datasets.
• The study confirmed the generality and scalability of the MBPR model.

Conclusions:

• MBPR offers a significant advancement in personalized medical service recommendation systems.
• The model's ability to integrate multiple user actions enhances recommendation accuracy.
• MBPR provides a scalable and effective solution for improving user experience on healthcare platforms.