Research And Improvenent Of Personalized Video Recommendation Algorithm Based On Item-based Collaborative Filtering

For the film industry’s prosperity and the rise of micro video, The video resources on the Internet is rising in an explosive way. Massive video resources solve the problem of the shortage of resources in the video but also brings a new problem -information overload. Users can not find their interested video from a large amount of resources. To exploring the interested video and recommended to the user from the massive video resources and meet the individual needs of users has become a new challenge of video portal site. Recommendation algorithm is a method of mining knowledge from the massive data, Different from search engines and it does not require the user to provide a clear demand but through analysing the user behavior mining user interest and recommends the items, which one meets the user’s interest, to the user. the traditional recommendation algorithm based on the characteristics,the recommendation algorithm based on user content and collaborative filtering can be used for video recommendation system, has the low degree of personalization recommendation, compared with the item based collaborative filtering algorithm.The accuracy of video recommendation and personalized degree is directly influenced by the video similarity and the accuracy user preference calculation. this paper combined with the characteristics of video recommendation and user viewing behavior preference of item based collaborative filtering recommendation algorithm, so that it can be better applied to the video recommendation system. This paper is mainly to improve the collaborative filtering algorithm based on item in the following two aspects:1. Analysing the influence of video similarity by simple statistics method to calculate co-occurrence matrix, proposed a method of co-occurrence matrix calculation based on the classification rules, calculate the video co-occurrence matrix more reasonable, improve the calculation accuracy of the video similarity.2. Modify the user preference formula, use the average score on the video of the user in the same group as the video quality factor and the users of the video type preference weight of users’ clear interest join to user preference formula.Finally, this parper use this improved collaborative filtering video recommendation algorithm make an experiment with the MovieLens data set, to determine the relevant parameters. And use the traditional item-based collaborative filtering algorithm as a comparative experiment, to verify the performance of the improved algorithm.