Web usage prediction and recommendation using web session clustering

In recent years, a strong interest has been given to web usage prediction and recommendation methods to improve e-commerce, search engines and other online applications. There have been various efforts carried out in this field, particularly focused on using recordings of web user interactions with websites. In this context, our research focuses on developing a novel approach for web prediction and recommendation. The proposed method relies on hierarchical session clustering by sequence similarity measure and takes advantage of access activity time and access position in prediction session to make a recommendation. The performed experiments reveal that hierarchical parameter and prediction accuracy are relevant. In addition, the paper introduces cost estimation to adapt web visitor behavior to web business purposes using prediction ansd recommendation results.