This paper focuses on users’ behavior towards an EC website. A novel Markov Chain-based way combining the web log file information and the topology of an EC website is presented to rank a user's interest in a WebPage. Then a URL-USERID relevant matrix is set up, with URL taken as a row and USERID as column, and each element’s value is the probability of a user to access a WebPage when time goes infinitely. The similarity of each column vector can be used to cluster customers, and relevant web pages can be found from the similarity of each row vector. The knowledge discovered by this dynamic model can be fairly helpful to the design and maintenance of a website, to provide personalized service, and can be used in an effective recommending system of an EC website etc.
Deng, Changshou; Zheng, Pie; Yang, Yanling; and Zhao, Bingyan, "Markov Chain-based Clustering Analysis of Customers and WebPages" (2004). ICEB 2004 Proceedings. 35.