Research On The Construction Of User Interest Model In Personalized Retrieval System

With the advent of the era of big data, how to manage and use vast data which already existed in the Internet is becoming a hot spot of the present. Each user generated a large number of historical data by the use of the Internet product, but normal search engines do not use these data, and the search results are usually unable to meet the individual needs of users. Therefore, personalized information retrieval system has become a research focus of new technology in internet.This paper introduces the inquiry-the concept of a bipartite graph theory, by analyzing the problem of the unreasonable weights on query design, put forward to divide user browsing history by query. In this paper, the query described by the concept which extracted from the user query history, when calculating the weights of query reference in the calculation of the query weights made reference to the query using time, the sequence of factors in the calculation of the query weights. In the process of user interests modeling, similar queries will be combined together by the clustering algorithm, and this reduce the dimension of user interest model, makes the proportion of each query in the model more reasonable. In storage and update the problem from model, this paper proposes storage format and update algorithm.Finally, using the construction of personalized retrieval prototype system, and capture user browsing history it, through simulation users browse verified query as the description of user interests is more reasonable. In contrast with other modeling algorithms the modeling algorithm proposed in this paper can get more excellent interest model. System operation results reflected the algorithm what designed in this paper is more reasonable.