Identifying Purchase-Stage by Applying Pattern Mining to Web Access Logs

We propose a method of identifying purchase-stage to enhance the success rate of recommendations. When a user purchases a commodity through the Web, she goes through several stages, for example, " Interest in commodity, Search for commodity, Comparison of commodities, Action of purchasing". To increase the purchase rate, we need to identify the different purchase stages and construct web sites so that they better support each purchase-stage. This paper hypothesizes that the purchase-stage of the user is evidenced by his/her web-access pattern. An experiment is conducted to collect the Web access logs of several subjects when they purchase commodities on electronic commerce (EC) sites. The proposed method is shown to approximately discriminate the purchase-stages.