A Novel Adaptive Video Recommendation Algorithm Under the Influence of COVID-19 Epidemic Situation

Objective: Video is a helpful way to spread information and culture. In the era of information explosion, it is challenging for users to find interesting and valuable information in massive videos rapidly. As valuable tools to solve information overload, recommendation algorithms can provide users with exciting information from a large amount of data. However, the traditional recommendation algorithms either recommend items according to the user's attribute information (e.g., content-based recommendation algorithm) or according to the user's history score (e.g., collaborative filtering algorithm). Without considering the COVID-19 epidemic situation, how to recommend effective prevention and control of epidemic knowledge. Users who hardly pay attention to the new coronavirus information need to recommend the necessary prevention and control knowledge and the latest epidemic development. For users who often view the relevant videos, it is also needed to control the recommendation list's contents to ensure diversity. Methods: To solve the above problem, we propose a novel adaptive recommendation algorithm that can consider user preferences and the epidemic situation's development and recommend valuable videos to users. Results: We have tested on four different datasets, and we choose two classic recommendation algorithms, namely user-based collaborative filtering algorithm and item-based collaborative filtering algorithm, to compare with our adaptive recommendation algorithm. The results show that, compared with the traditional algorithm, our algorithm can not only recommend videos that meet the user's taste but also consider the transmission of widespread science knowledge in the case of an epidemic situation. The recommended video information can be dynamically adjusted according to the user characteristics to ensure the recommendation list's diversity. Conclusion: We have conducted extensive experiments on real-world datasets. Compared with state-of- the- art recommendation algorithms, the recommendation results can match user preferences and give reasonable consideration to the epidemic's information, demonstrating our algorithm's effectiveness.