Web Log Mining, Research And Realization

As the Internet grows, the sheer volume of information available on the Internet is overwhelming. This phenomenon is referred as information overload. The information diversity in the Internet makes it even harder for users to find the desired information. Users are lack of effective ways of retrieving relevant information and are easily got lost in the cyberspace, namely information bewilderment. Nowadays, we primarily use search engines for information retrieval. Most search engines, however, perform passive searching and return results regardless of user preference or accounting for no user specific interests. Therefore, search engines themselves cannot effectively solve the information overload and information bewilderment problems.Among many direct and indirect solutions, employing data mining techniques on web log is a promising approach. By identifying the browsing pattern of users based on their logged browsing activities, we design web services that integrate user browsing pattern so that the website is personalized and becomes intelligent.In this thesis, we systematically introduce the entire process of data mining, web data mining and web log data mining. We review the current applications of pattern discovery algorithms and discuss the problems that they have. Based on the support-preference concept, we designed a website access matrix and a new path mining algorithm (PMA). Experiments demonstrated that the new PMA accounts for the browsing preferences of users accurately and exhibits good extendibility. We designed and implemented a web log data mining system (WLDMS) and applied it to a campus website. We performed data mining on web log and used acquired user behavior patterns to construct an intelligent website.