The Research Of The Hybrid Model In The Personalized Recommendation

With the rapid development of the Internet and the explosive growth of information, the information overload problem is getting worse; this phenomenon results in people cannot get the information which they need in a short time. A method to solve the information overload problem is recommendation. Therefore, recommendation system will come into being. When recommendations appeared firstly, they were specifically targeted at the specific issues. Contents of the recommendation are for users to filter messages. Over time, people in their daily online shopping, watching movies, listening to music when they hope very easy to get the information which they want. And personalized recommendation also appeared. Personalized recommendation is to predict the behavior of users for possible future based on the user’s personal preferences. Each personalized recommendation algorithm has its own advantages and disadvantages, in the plurality of information today. There is not a way to solve all these problems.In this paper, It based on the analysis on recommendation algorithm’s advantages and disadvantages. In order to solve cold start and the data sparseness problem it proposed a hybrid personalized recommendation model. Before run the recommendation algorithm, the data set will be deal with by clustering algorithm. It can reduce the data set. In this paper, we modify the core part of the clustering algorithm which calculate the distance. We apply with Hamming distance in the clustering algorithm. When we considering the similarity between users,we added the relationship between users and items. When we considering the similarity between items, we added the relationships between items and users. This clustering method, can increase the accuracy.Personalized recommendation of the hybrid model is designed in this paper is divided into three modules. The first module is a social recommendation module. The social recommendation module is mainly used to solve the cold start problems. If the new user account is associated by a social account. Then the user can recommend something by his friends. The second module is recommended based Slope one module. It worked without cold start problem. It worked as the main part of the algorithm. The third module is a content recommendation module. It worked with new item. It can recommend new item to user.In the last part of the paper, Movie Lens is used to verify mixed models and movie recommendation system. In the end, it give analysis of the effectiveness,legitimacy and effectiveness of the application of the model.