Mining Users' Interests Based On Search Logs

Search is valuable in the Internet industry.The user's true intent resides in the search log generated by the user's interaction with the search engine.Search log data generally contains user basic information,user query words and user-visited URLs.The user query terms focus on the user's interest tendency,while the user's web page text contains more user interest information.Through the analysis and mining of massive search log data and web page data captured by search engines,users' interest tendency can be obtained.Mastering the user's interest tendency can provide better search service for users and find potential user value.The main content of this paper is how to mine the user's interest through the search log.This paper starts with massive search logs and uses statistical knowledge to generate a summary of search log data and data preprocessing of search log data.The user interest model is constructed,and the user interest vector is used to describe the user,and the user similarity evaluation index is generated.Through clustering algorithm,the model is constructed to cluster the user and mine the interest of the individual cluster users.When mining the user's interest in browsing web pages,this paper constructs a number of text classification models using annotated web page text data.The model is evaluated by multiple indexes,and the optimal model is selected.In order to construct the user clustering model and text classification model,this paper compares several different text representation model,practice and compared a lot of text processing techniques and methods,including Chinese word segmentation,word vector technology,word reduction technology,feature selection method etc.