Study of Markov Chain-Based Dynamic Customer Clustering of Users on EC Website

A systematic investigation into users´ browsing behavior towards an EC Web site was made. A novel Markov chain-based way combining the Web log file information and the topology of an EC Web site was presented to rank a user´s interest in a WebPage. Then a URL-USERID relevant matrix was set up, with URL taken as a row and USERID as column, and each element´s value was the probability of a user to access a WebPage when time goes infinitely. Finally, using the URL-USERID relevant matrix, a clustering matrix was proposed to group similar users. The clustering results are fairly helpful to providing personalized service and implementation of customer relationship management