Discovering user's interest at E-commerce site using clickstream data

Abstract

It is a common view that users' browsing behaviors reflect their true interest in commodities at an e-commerce site. With the development of E-commerce, users' detailed navigating and purchasing behaviors can be completely stored. Clickstream data are records of users' browsing behaviors, which provide information about the path viewed by users and their access time on each page. Usually, personalized services can be offered based on this information by e-commerce service providers. To facilitate dynamic and personalized commodity recommendations, not only within-category but also across-category, an interest oriented method based on clickstream data mining is proposed in this paper to cluster users. A new definition of users' interest is introduced for the first time as a set of the preference for commodity categories. In order to describe users' behaviors and reflect their interest, three main indicators category visiting path, browsing frequency and relative length of access time are taken into consideration and refined from clickstream data. According to these indicators, an improved clustering algorithm with rough set theory is used to cluster users with similar interest. The experimental result shows that this algorithm is effective and applicable. The result of this proposed algorithm can be applied to support decision making for e-commerce sites.