Mining Web Logs With Association Rule Incremental Updating Algorithm

The rapid improvement of Internet and World Wide Web makes the design and maintenance of web sites more and more important. In order to make the users able to conveniently browse the web sites, it is necessary to thoroughly analyze the users usage information of web sites and construct more rational web site structures. Especial for E-business web sites, through discovering the browse rules of customers to provide they more personalized content even relates to the survive or perish of this sites. Applying data mining technology to large-scale web data, web mining can discover potential patterns about customers browsing behavior, which has great promise.Focusing on the E-business aspect of web mining, this paper adopted association rule mining technology and proposed a complete resolve method. This paper proposed new opinions and methods about the process of data preparing,mine algorithm and personalized recommendation.Firstly, this paper completely analyzed and discussed the technologies of data mining,association rule mining,association rule incremental updating algorithm,web log mining algorithm and the structure of web log mining. Secondly, aiming at the characteristics of E-business, this paper studied the data cleaning step of data pre-processing and proposed new method of data cleaning which can improve the efficiency of association rule mining algorithm. Then, in order to resolve the problem of network delay, this paper studied the transaction recognition step of data preprocessing and proposed an improved time-window based transaction recognition method. Further more, in order to satisfy the needs of association rule mining algorithm in E-business field, this paper studied the support-ordered tree based association rule mining algorithm FOLDARM and proposed the concept of sequential support-ordered tree. This paper used the concept of mothball frequent item sets for reference to make use of historical mine results in greater extent. In corresponding, this paper improved the tree structure with mothball frequent item sets and proposed new item updating algorithm.. At last, this paper proposed new methods of recommending web pages to indicate the purpose of web site's managers.