Hybrid Recommender System for Personalized Patient Care

Abstract

This study presents the development of a hybrid recommender system designed to support clinical decision-making in healthcare. The system integrates four distinct approaches— collaborative filtering with matrix factorization, neural collaborative filtering, content-based filtering, and knowledge-based filtering—within a unified neural network framework. By leveraging both patient-specific data and domain knowledge, the system provides personalized recommendations for medical procedures. Experimental results on the MIMIC-IV dataset demonstrate the effectiveness of the proposed model, achieving a Mean Absolute Error (MAE) of 0.0842 and a Mean Absolute Percentage Error (MAPE) of 8.42% on the test set. These outcomes highlight the huge upside potential to be applied such as tailoring services to patient profiles, optimizing resource allocation, and supporting targeted healthcare promotion. These results demonstrate the practical value of hybrid recommendation models in bridging data- driven insights with medical domain knowledge.