Discovery of Significant Usage Patterns from Clusters of Clickstream Data
Lin Tao Lü, Margaret H. Dunham, Meng Yu · 2005
Discovery of usage patterns from Web data is one of the primary purposes for Web Usage Mining. In this paper, a variation of “user preferred navigational trail ” called Significant Usage Pattern (SUP) is proposed. SUPs are patterns that are extracted from clustered abstracted clickstream data, with a higher normalized probability of occurrence and may begin/end with specific Web page(s). The novelty of our approach is in the application of clustering to data abstraction based on a new twophase abstraction technique. In order to generate SUPs, first, the Needleman-Wunsch global alignment algorithm is applied to the sub-abstracted sessionized clickstream data to compute the similarities between each pair of sessions. Based on pair-wise alignment results, a similarity matrix is constructed and then sessions are grouped into clusters according to their similarities. Web sessions are abstracted again using a concept-based abstraction approach and then a first order Markov model is built for each cluster of sessions. The specific navigation paths, i.e. SUPs, with a normalized product of probability along the path above a certain threshold and beginning/ending with specific states are generated from each cluster based on its corresponding Markov model. Experiments conducted using Web log data provided by J.C.Penney show that different clusters of Web sessions may generate very different SUPs.