Research And Design On Personalized Books Recommendation System

At present, various online books recommendation systems have been put into use, used to help users quickly and easily select books they need. But the low individuation degree is a widespread problem of online books recommendation systems. With the way online bookstores used more and more being accepted by users, and differences of services each online bookstore provided to users are also more and more small, personalized recommendation technology can increase the competitive advantage of online bookstore and has broad application prospects. A lot of research on online books recommendation system, mainly concentrated in the information acquisition modes and recommendation technology. In information acquisition modes, how to effectively combine all sorts of scores information, and to more fully reflect users'needs are concerning problems. In the research of recommendation technology, how to determine the recommendation algorithm is a key problem. At present many online bookstores are using collaborative filtering technology. In this technology, how to define users'similarity and how to select the user reference group is the focus of research on collaborative filtering algorithm. And the difficulties in it include data sparseness and algorithm scalability problems.In view of the above question, this paper studied the following several aspects:1. It studies a method of applying content-based filtering and cooperative filtering technology in the personalized books recommendation system, and the effective combination of the two techniques used in this system. Introducing resource attribute similarity and clustering method, it studies content-based filtering; It focused on the research of method combined with explicit user scores, recessive user scores and predict user scores and the method how to form nearest neighbors in the cooperative filtering.2. It uses comprehensive interest score method, which is a combination of explicit user scores, recessive user scores and predictive user scores, to solve the sparseness and oneness problem of users'interest scores in the cooperative filtering, in order to form a more accurate and complete user interest score. On the user recessive scores, this paper determines the user's interest with two factors:clicks and browsing time. On the predict user scores, using the result of similarity measurement of attribute vector,and according to the similarities each user's having scored books class and candidate books class to fill the items not been scored in the user-book class evaluation matrix.3. It analyzes and designs a personalized books recommendation system, and implements it by a personalized books recommendation prototype system. The explicit process of the recommendation module is illustrated by a concrete application case. Finally it conducts an experiment on the prototype system, and the experimental results show that the personalized books recommendation system can respond to the needs of users, and play some recommendation effect.