Effective Personalized Web Access Pattern Mining Method Based On Fuzzy Clustering

Extracting patterns from web usage data can help to achieve better web personalization and web structure readjustment.Currently,the Web Log mining is still in the exploratory stage,mature theory and methods have not been fully formed and needs to be further studied.Traditional Association Rules algorithm cannot deal with increasingly large amounts of data and variety of data types,and mining of web traversal patterns based on frequency can not provide sufficient information for terminal users.This paper uses a web traversal mining algorithm based on rough clustering for group web traversal patterns from web log.First,in order to metric similarity and difference between any two web access patterns,the fuzzy vector with same length is used to describe the browsing time for web user,each of these elements is either 0 or fuzzy vector represented by fuzzy language.It not only indicate the user visited the web page,but also can describe the web browsing time.The value of the browsing time is represented by fuzzy linguistic variables,it's in accord with the people's way of thinking and we can ignore the subtle differences between any two access time.Because of the clusters may not exist crisp boundaries,thus a fuzzy rough k-means clustering algorithm is proposed to cluster these fuzzy vector which representing the user browsing features.Experiments show that clustering method can reduce the complexity and increase accuracy.Secondly,considering both the web access frequency and web access time can reflect users' interest,we propose a new method to extract web traversal patterns.This method efficiently extracts web traversal patterns based on clustering,an evaluation function defined by the user is adopted to calculate the value of the patterns.According to the experimental results of different datasets,this method can effectively extract web traversal patterns from large-scale datasets.