| Because of the rapid integration of video service platform and social tool, users can easily share their interests with friends on video service platform. At the same time, users need to spend a high price to get the content of interest among massive video resources.Therefore, by analyzing the historical behavior of users, then design a recommendation system based on user interest preferences model in order to provide users with personalized family service initiative, which has great research value.The user's interest is dynamically changed over time, but the traditional recommendation algorithms ignore the influence of time. To this end, the time weight method of calculating the label weight is adopted in this paper, and according to the user's preference, the paper further divide the user's interest. With the users of video service platform increased gradually, adopting the method of user clustering which will assign the similar users to the same user group to solve scalability problems. In addition, combine the content-based recommendation algorithm and collaborative filtering recommendation algorithm together to provide better personalized service. The recommendation system just passively analyzes the user's history, so the paper predicts the popularity of the video initiatively on the service platform, which can guide recommendation system for customer service, as well as solve the cold start problem of recommendation.In order to have a real experiment environment, add the WeChat Public Number platform and mobile client platform on the original findball website. Experimental results show that the recommendation algorithm in this paper has greatly improved the accuracy when compared with the existing methods. In practice, the paper use different caching strategies to reduce network response time between server layer and client layer, To improve the real-time performance of personalized services, our future work will focus on designing a distributed index structure for big data. |
PDF Full Text Request