The Design And Implementation Of Personalized Recommender System Based On The Analysis Of User’s Hidden Behaviors

With the rapid spread of Internet, human society is going enter "information overload" era. How the users can quickly find relevant information from the huge amount of internet data? The question has become a popular topic of internet technology. Personalized recommendation is a new kind of information filtering technology. It can find users’interest preference from users’historical data, then give the information to user with the way of "push". For electronic commerce website, this can improve the user experience and maximize the benefits.Through analysis and research the collaboration filter algorithm in the personalized recommendation system, this paper proposes some improvement methods to solve the existing problems of collaborative filtering such as data sparseness, data accuracy, cold start and so on. Among the problems, sparseness of the data and cold start seriously influence the accuracy of recommendation.(l)Using distribution processing platform to analyze the users’hidden behaviors data, establishing interest model of users, can effectively solve the problem of data sparseness.(2)While cluster the users, users in the same class have projects of preference of partial similarity. In the process of cluster, identify the project of preference of partial similarity. When doing personalized recommendation, we only need to calculate the user similarity of these projects of preference of partial similarity instead of all the projects. This method can effectively improve the flexibility of algorithm.(3)To solve the "cold start" problem, we use the means based on Professor and machine learning. The experts summarize the feature gene library which generated by the feature information of all the projects, and then generate the project feature gene of each projects, generate user interest feature gene for "new users". Through techniques of nature language processing, automatically renew and improve the projects’gene library constantly and solve the "cold start" problem through gene similarity.Finally, according to the characteristics of e-commerce systems, we design a e-commerce personalized recommendation system which is combined with e-commerce website and personalized recommendation system based on users’ hidden behavior. We verify the system on the real data set and look ahead the defects of the system.