A User-Based Collaborative Filtering Algorithm Considering Interest Attenuation for Short Video Recommendation

To solve the problems of recommendation delay and repeated recommendation in short video recommendation, this paper propose a user-based collaborative filtering algorithm considering interest attenuation for short video recommendation. First, we extracted the user's positive feedback behavior from the user's historical interaction behavior records and quantified the user's interest through weighted calculation. Then, we build an interest time decay function by taking into account the features of user interest attenuation, which in turn improves the methods of similarity calculation and rating prediction in the collaborative filtering algorithm. The offline experimental results show that the algorithm we proposed outperforms the traditional algorithm in accuracy, proving the effectiveness of the improved method and providing implications for solving the problem of users' changing interest.