Aggregated clustering for grouping of users based on web page navigation behaviour

Abstract

In this epoch, a significant amount of patterns are retrieved using data mining techniques. Clustering is one of the technique that plays an vital role in web mining. This paper works on MSNBC dataset with the average access length of 6. It aims to cluster the users based on their navigation behaviour. An iterative aggregated clustering is proposed, in which various clustering algorithms like EM clustering, farthest first, K-means clustering, density based cluster, filtered cluster are applied on the dataset. The resultant clusters from various algorithms are aggregated correspondingly and the frequency of instances in each cluster is determined. Then the instance with two-third majority is grouped in that cluster. The work revealed that 91% of users clustered in the first iteration under 17 clusters and 99% of users in subsequent iterations in another 17 clusters and rest of the users are grouped as one cluster, resulting 35 hard clusters.