The Research Of User Intention Analysis Of Based On User Behavior

With the proliferation of information on the web, also the types of information and the number of the users and their roles, how to meet people's information need fast, accurately and comprehensively is a large challenge that we need to face. General search engine is based on the keyword matching technology, due to the short query the user entered and the ambiguity of the query. The retrieval results are far from satisfaction. Mainly in the following points: first, there are many irrelevant documents in the search results .Most of them that are not relevant to the user intention, although they contained the keyword. The information on the Internet is in a dynamic state and each search engine has its update cycle, if the user wants to know the latest information about this field, they have to resubmit the query after a few time. Then the adaptableness of the search result, if different users submit the same query to the search engine, the same results will be returned without considering the differences between different users.These problems cause the low accuracy of search results on one hand, on another hand users have to spend much more time to find their interest topics. So the technologies of query optimization and query expansion are put forward in order to locate information request of user accurately. So how to understand the user's intention and provide personalized services is very important.The main contents of this paper are summarized as follows:First, this paper presents a clustering algorithm based on user's search habit and preference. After this, the users with the same interest are clustered into one cluster, named a community and the interest field they are interested in is obtained. Through this method we can mine the potential intention of the user.Second, the clustering method is transplanted to the literature retrieval system. Based on the above clustering results, we can provide the personalized service closer to the user intention by the recommendation of the users with the same interest. The process contains two steps: objective recommendation and subjective recommendation.Finally, we conduct the experiment to validate the result and take the literature retrieval system an example. For the same query, we collected the user search behaviors according to some rules and identified the user with the same interest using the clustering algorithm. And then HITS algorithm and the collaborative filtering technology are applied into the literature retrieval system. And the experiment shows user can get the personalized service by the similar users'recommendation.