ABSTRACT

Web usage mining is the art of discovering navigation patterns of users from web log data. Next web page prediction, a task of web usage mining, is used to envisage future requirements of the user during surfing. This paper presents an ensemble prediction system, that uses ensemble clustering and ensemble classification. The proposed system uses heterogeneous clustering ensemble model to group similar browsing sequences together, which is then used by a heterogeneous classification ensemble model to predict future requests of the user. The goal of this combination process is to improve the quality of individual data clustering and classification. Experimental results demonstrate that the combination of ensemble is efficient in terms of prediction accuracy and can be used by web masters to attract users.