Research on Recommendation Algorithm for Short Video Health Content Targeting Elderly Users

To achieve precise delivery of health-related short videos to elderly users, this paper proposes a real-time recommendation method based on a knowledge graph. This method integrates both short-term and long-term interests of elderly users, designing an efficient algorithm to capture and streamline short-term interests. By leveraging users’ recent interactive behaviors and the node relationships within the knowledge graph, it rapidly extracts health topics of concern. Additionally, an interest fusion mechanism is developed, which combines long-term interests (extracted through knowledge graph mining) with short-term interests (obtained through streamlined representation). Weighted summation or an attention mechanism is employed to dynamically adjust the weights, enabling personalized recommendations and enhancing efficiency and accuracy. Experimental results indicate that the accuracy and Mean Reciprocal Rank (MRR) of the proposed algorithm are $85.19 \%$ and $85.09 \%$, respectively, demonstrating its high precision. Moreover, the algorithm maintains the highest F1 score and accuracy across all predefined recommendation lengths.