implementing user clustering using web log data

The sophistication of the web browsing depends on the likelihood of user reaching the intended web page at the earliest. Patterns of usage have to be tracked in order to conclude the user’s intent. These usage patterns are analyzed with respect to URL’s identified from user segmentation which is obtained from finding similar behavioral patterns. The technique employed here to achieve this phenomenon is the formation of user communities using pattern mining technique. Agglomerative clustering methods are used to typically discover the strongly linked communities. But these methods prove to be inefficient when implementing in large networks with several thousand nodes. Hence the method of triad formation, with graph mining can be applied to resolve new combination of edges which are not connected and establish links between them thereby increasing the identification of active users. The methodology is to form links between URL’s that are most frequently used. The prerequisite for this is to detect communities based on the user’s interests and increasing connectivity. The ultimate aim is to help the network providers serve their customers based on the user interest. The genetic programming method of churn prediction is applied to identify the possible customers who may leave the network with the help of which steps to retain the existing customers in a network can be made