Fusion User Difference Degree And Information Entropy Of Collaborative Filtering Recommendation Algorithm

With the rapid development of Internet, the mass of information is presented around us and constantly update, that makes selection difficulties, users cannot to find needed information in vast amounts of information, this problem is known as information overload. In order to solve the problem of "information overload", all kinds of recommendation system arises at the historic moment, collaborative filtering recommendation algorithm is the most mature technology, but at the same time some collaborative filtering also exposed the shortcomings and defects: data sparse, cold start, etc. How to solve these problems has become the hot topics in the study of the current recommendation system.This paper first introduces the related research achievements of collaborative filtering algorithm in recent years, and recommended system related technology, then introduces the commonly used several kinds of recommendation algorithm, this paper mainly introduces collaborative filtering recommendation algorithm and the similarity calculation method and the existing problems, and then analyze the shortcomings of collaborative filtering algorithm, the classic collaborative filtering algorithm is improved. Finally in this article, through the simulation experiments, the recommendation algorithm can improve the testing and analysis, illustrates the feasibility and effectiveness of the improved algorithm.Jaccard similarity calculation is presented in this paper, on the basis of combining with the users of the project of a grade difference degree and information entropy to improve the similarity calculation method, put forward two kinds of improved user similarity measure method. Finally in order to verify the effectiveness of the similarity calculation method, by experiments on Movie Lens dataset, shows the application of the proposed recommendation system similarity algorithm of recommendation effect are improved significantly...