A New Similarity Measure for Generalized Web Session Clustering

A new similarity measure for generalized web session clustering is defined on the common paths of users´ navigation patterns (SMCP) in this paper. It divides the similarity of common paths between two sessions into the inner part and the outer part. The traditional k-means is employed to test the performance of similarity measure and by comparing with the visiting order Based (VOB) similarity measure and the path angles Based (PAB) similarity measure, the experiments on both real and synthetic datasets show that clustering using the proposed similarity measure SMCP can yield more than 10% higher accuracy than the VOB and PAB similarity measure in terms of Silhouette value.