Collaborative Filtering In Personalized Recommendation Based On High-dimensional Subspace Clustering

In recent years, with the rapid development of the Internet and information technology the problem of information overloading and information amazing, which we are having been frustrated, has became more and more worse. And all these boost the flourishing development of personalized recommendation systems. Personalization recommendation systems were proposed to provide the target user with information they interested in based on the existing information, so users can be more convenient to find the information they need. Recommended System in has good prospects for the development and application, gradually become an important research, which has been more and more attention.Collaborative filtering recommendation (CFR) algorithm can make choices based on the opinions of other people. It is the most successful technology for building recommender systems to date. The basic idea is to found a good way for users to find the contents of his genuine interest is to find him and his interest in similar users first. Then these users interested in the content of recommended to the user. The tremendous growth in the amount of available information and the kinds of commodities to web sites poses some key challenges for recommender systems, so a series of problems in collaborative filtering recommendation are needed to be solved. In this dissertation a profound analysis of these problems are given. The main points are described as follows:Firstly, this dissertation analyzed the traditional user-based and item-based collaborative filtering algorithms, researched the design of algorithms for collaborative filtering and some key technologies of personalized recommendation systems. Through analyzing the problems of the modified collaborative filtering algorithms, this dissertation explored and researched affect the quality of the sparse recommended problems and issues affect user satisfaction with the recommendation of integrity in collaborative filtering.Secondly, fort the sparsity and dimension disadvantage of high-dimensional item- score data, subspace clustering algorithm of high dimensional data was introduced. Clustering are produced by subspace clustering algorithm based on users pattern similarity, and collaborative filtering algorithm was improved by calculating of model similarity which brings recommendation to users. At the same time, a off-line clustering method is proposed, and it increased the response speed of the system evidently. The experimental results show presented in this dissertation based on users-pattern similarity subspace clustering algorithm, and on this basis, the results of the collaborative filtering recommendation is superior to the traditional K-means clustering algorithm based on the recommendation of results, especially in light of the evaluation of user data sets reflect on the recommendation of a good performance.Lastly, based on the collaborative filtering in personalized recommendation based on users pattern subspace clustering. As well, the next simulation experiment with item factors demonstrate that the proposed collaborative filtering recommendation improved algorithm based on user patterns clustering and item sets in the accurately, integrity, diversity and so on are all further improved while keeping the temporal ability.