A Personalized Recommendation System Based On Hybrid Algorithm

With the popularization of PC and smart phones, the Internet has covered all over the word, and people get information from the web more conveniently. However, in addition to the information we need on the Internet, there are a lot of noise information. And we need a lot of time and effort to filter them. How to deny information to the Internet, that people can quickly find what they want is more and more important.Personalized recommendation system has been put forward, so that the Internet can provide people with personalized information services. The recommendation system engine - recommended technology plays a crucial role in the recommended effect of the recommended system. A good recommendation engine can make people a surprise, a poor recommendation engine is like garbage generator. Therefore, how to improve the performance of the proposed algorithm becomes an urgent problem to be solved.In this thesis, on the basis of the existing recommended technology,a hybrid recommendation algorithm based on user active to weight is proposed to improve the scalability, stability and accuracy of the recommendation system. The main work and contributions are as follows:1. A new scheme of collaborative filtering algorithm - collaborative filtering based on label weighting has been proposed. In this paper, we propose the method of user label weighting, which effectively reduces the amount of on - line computation and improves the scalability of the traditional collaborative filtering algorithm.2, Apply FFM model to the recommended areas. Using the appropriate loss function and optimization method, the FFM model is applied to the score prediction, which solves the problem of the poor performance of the collaborative filtering algorithm in the sparse data set and the new users.3. Put forward the idea of determining the mixing coefficient based on the user's history activity, and determined the mixed coefficient formula according to the experiment. In this paper, a new recommendation algorithm is proposed by mixing the weighted-based collaborative filtering algorithm and the recommendation algorithm based on FFM model.4. Designed and implemented a personalized book recommendation system based on the hybrid algorithm which has been proposed in this paper. The function of system includes user registration and login,information management, online real-time recommendation and so on.