Proposed Partitioning Approach and Hierarchical Approach for theClustering of Web Users

摘要

In web usage mining, the cluster analysis is the most important technical. Based on this technical,the users groups provide insight into browsers behavior to refine the behavioral patterns to the website accessand to identify frequently visited pages. Into the clustering context, the most popular approach is thepartitioning approach but itprinciple as it stands seems inappropriate on sequential data. This work attemptsto over come limitations and proposes a new model of users clustering. This approach is based on setofmeasures which ensure rules quality and evaluate the interest of generated sequential rules. The experimentalstudy implements the proposed algorithm, the k-medoids algorithm and hierarchical agglomerative approachin order to guarantee the good partitioning of the data interms of evaluation measures of the clustering qualityand calculation time. Relative result sattempt to ensure a good partitioning data in terms of evaluation measuresof clustering quality.