Clustering Web Log Data Using Graph Partitioning and Agglomerative Hierarchical Algorithms for Predicting User Navigation Patterns

Abstract

Web usage mining enables organizations and website owners to study the user access patterns or behaviors when navigating their websites. In predicting user navigation patterns, previous study has proposed a two-stage clustering-and-classification of web log data. The main issues in web usage mining is the precision of recommendations in user navigation patterns since it’s will affect the quality of prediction of user future navigation. The quality and precision of navigation patterns produce in the clustering stage is useful contribution in designing an accurate user prediction system. This paper aims to undertake comparative analysis on two clustering algorithms, which are graph partitioning and agglomerative hierarchical clustering to compare user navigation patterns produced by each clustering technique. The results from clustering experiments showed that graph partitioning algorithm produced a detailed list of navigation patterns compared to agglomerative hierarchical clustering.