Dynamic recommender system for chronic disease-focused online health community
Junruo Gao,Yuan Zhao,Dongming Yang,Yuyang Liu,Liang Zhao
DOI: https://doi.org/10.1016/j.eswa.2024.125086
IF: 8.5
2024-09-02
Expert Systems with Applications
Abstract:Unequal distribution of healthcare resources poses a significant challenge in numerous regions and countries worldwide. Online Health Communities (OHCs) serve as pivotal platforms for patients to share treatment experiences and seek emotional support from peers facing similar health challenges. The facilitation of efficient user navigation to relevant sub-communities and topics within OHCs emerges as a critical need. Addressing these complexities, a recommendation system specifically designed for OHCs is proposed to tackle unique challenges such as simultaneous prediction of sub-communities and topics with varying levels of information, capturing the influence of evolving health stages on patient interactions, and ensuring coherent recommendations. To meet these challenges, OHCs are conceptualized as dynamic tripartite graphs that model three primary entities—patients, sub-communities, and topics—alongside their temporal interactions. Novel techniques including dynamic tripartite subgraph convolution and dynamic tripartite collaborative filtering are introduced to effectively capture the impact of changing health stages. Additionally, a multi-type entity prediction method is developed to forecast sub-communities and topics, incorporating robust constraints to ensure the validity of predictions. The study conducts extensive experiments using real-world datasets from two online chronic disease communities, including the Breast OHC dataset and the Bladder OHC dataset, encompassing 5,346,591 and 16,427 patient interactions, respectively. Results demonstrate significant advancements over state-of-the-art methods, achieving maximum improvements of 11% and 15% in Mean Reciprocal Rank and Recall metrics, respectively. Sensitivity analyses of online and batch-based algorithms provide insights into the trade-offs between efficiency and accuracy, supported by an illustrative application example highlighting the system's superiority over existing approaches.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science