Research on Personalized Recommendation System of New Media Short Video Based on Local Random Walk

With the rapid expansion of network information and resources, users are faced with the difficulty of screening the information they need from massive data. It is necessary to find an effective tool or method to help users complete this information screening process. The core of the new media short video recommendation system is the short video recommendation algorithm, which predicts users' tendency to short videos based on their behavior records and the characteristics of short videos. Under this background, personalized recommendation algorithm came into being. Aiming at the sparseness and cold start of existing recommendation methods, a short video recommendation method is generated by random walk based on short video tags. Random walk is the research method adopted in this paper. It controls the depth of random walk and tries to select the rating information of the neighboring users who are close to the target users in the trust network. According to the research in this paper, the recommendation results generated by random walk method have better ranking quality than traditional algorithm data sets. Generally speaking, the ranking quality of random walk is higher than other methods.