Technology And Application Of Personalized Recommender Systems

With the popularity of Internet and rapid development of E-Commerce, the resources in the net increases exponentially. Due to the huge amount of resources, the phenomenons of"information explosion"and"information overload"arise. Personalized Recommender Systems emerge under the background of this, which becomes the research focus in the domestic and overseas. In the past twenty years, recommender systems have made significant progress and become one of the most successful tools to solve the"information overload"problem. However, far from perfect point, the research on personalized recommender technology and systems also remains in the initial stage. A lot of problems still exist to be solved.According to the different recommendation algorithms, this paper introduces four kinds of recommender systems: rule-based recommender systems, content-based recommender systems, collaborative filtering based recommender systems, and hybrid recommender systems. The main research works of this paper are as follows:(1) On the basis of a brief overview on the research of recommender systems at home and abroad, a summary on four kinds of recommendation algorithms is given.(2) Discuss the sparsity problem, cold start problem in the traditional user-based collaborative filtering algorithm. And then propose a novel collaborative filtering method based on user interests transmission(UIT). This method computes user similarities in the dimension of interests, and it considers the interests transmission between different users. To some extent, this method not only can cope with the high dimensionality and sparse data problems, but also have higher precision.(3) Put forward and design an improved association rule-based recommender system by dividing all the users into different interest groups. First, build the users'network according to similarity between user's behavior and intersts. Then, by running a community-discovery program on this constructed network, we can get a number of groups representing different interests. Then, mine frequent itemset in each group to obtain rules belonging to different groups. At last, provide personalized service to users by matching the historical behavior of the target user and frequent itemsets of the related group.