Access Log Analysis with PageRank

SUMMARY In recent years there has been an increasing demand for access log analysis for use in marketing, because log data contains information about user behavior. If a user's behavior can be acquired, information about the efficient optimization of Web sites can be provided. The main objective in this paper is to discover knowledge about customers' behaviors in the purchase process based on access log data on an electronic commerce site. We construct a network from the access log data and measure the importance of Web pages by analyzing the network using the PageRank algorithm. Since the network represents customers' transition information, we can evaluate the Web pages according to the customers' behavior. We confirm that the proposed method found rival products and differences in customers' behavior among product categories. We also confirmed that the ranks generated with the proposed method were different from those generated by comparable approaches, specifically the PV and sojourn time.