Application of Time-Weighted User Behavior Analysis Collaborative Filtering Algorithm in Short Video

Abstract

In the big data environment, the information overload is serious. The traditional search engine does not satisfy the user’s demand for information. The user wants to obtain the user’s acquired information conveniently and quickly, and the information producer wants to transmit the information to the target user. The intermediary that acts as both is the recommendation system. Recommendation system is an important way to deal with massive information. This technology brings great convenience to viewers in the field of short video. It is very difficult to extract features from missing short video data and from content information of short video. About this question. This paper proposes a time-weighted user behavior analysis collaborative filtering algorithm. This paper starts with the behavior of short video users. In order to compensate for date sparseness, the explicit and implicit behaviors of users are deeply analyzed and a user-video scoring model is established. The user’s interest in video will change with time, so we use time weighting to adjust the time variable. Finally, the K most similar neighbors are extracted and recommended to the user in order from high to low. Experiments show that the proposed algorithm alleviates sparsity of date and improves the accuracy of recommendation.