Multi-criteria Health Science Short Video Recommendation Inspired by Heuristic Optimization

An increasing number of people are obtaining healthcare and disease prevention-related knowledge through health science short videos, guiding their own health behavior habits in recent years. However, users may face the information overload problem in their searching process. Personalized recommendation can provide users with information related to their interaction history. Traditional personalized recommendation methods originate from e-commerce applications. Their excessive pursuit of click-through rates limits their usefulness over the health science short video platforms. Therefore, this study proposes a multi-criteria recommendation model that balances multiple recommendation goals. Specifically, we propose an item graph searching and updating encoding scheme that utilizes the relationships between discrete items to enhance the solution searching efficiency and the performance robustness. Results on real world data show that the proposed search scheme can effectively achieve multiple recommendation goals simultaneously.