Web Access Patterns Mining for Individuals with Timing and Link Sequence

摘要

The issue of user identity trustiness has become more and more serious in e-commerce recently. Researches on evaluating user identity trustiness by user behaviors have received great attentions recently. In this work, proposed a method of constructing the individual access patterns based on the relationship among web pages they have visited by web log mining. The web pages were clustered based on the contents and timing frequent sequence. Then, Individual Access Pattern (IAP) was proposed to describe the patterns of individual access behaviors. IAP stored highly compressed, critical information for individual access behavior and habits in a fine-grained way. IAP also could be easily transformed to Markov Model, which could be used to compute the confidence of various users. Thus, IAP could be applied to e-commerce transaction for identifying users in real time. The average accuracy of individual pattern recognition can reach 88.5% after the user browse 5 webpages. FAR (false accept rate) or FRR (false reject rate) can be lower than 10% by adjusting different strategies. The experimental results showed that IAP can reflect the differences of various individual access patterns both from the timing sequence and the hyperlink relationship precisely.