A Data Cube Model Supporting Web-log Mining
Takanobu Sukegawa, Tadashi Ohmori, Mamoru Hoshi, Yuichi Tsutatani · 2003
Frequent pattern detection from a large Web-log dataset has been considered a useful method of Web-site anal- ysis. However, a naive usage of frequent pattern detection may lead to producing another meaningless dataset, namely, too much or trivial frequent patterns about which Web-pages were visited. To overcome this difficultly , this paper proposes a new datacube model supporting interactive process of Web-log mining. This datacube is powerful when a human user continues Web-log analysis by using frequent pattern detection and under a traditional datacube approach. This paper describes our new datacube model, denoted by itemset cube, and then efficient data-processing algorithms of three associated operations materialize, roll-up and drill-down are described.