An Enhancement in Clustering for Sequential Pattern Mining through Neural Algorithm Using Web Logs

Abstract

An Organization need to understand their customers' behavior, preferences and future needs which depend upon past behavior. Web Usage Mining is an active research topic in which customers session clustering is done to understand the customers activities. This paper investigates the problem of mining frequent pattern and especially focuses on reducing the number of scans of the database and reflecting the importance of pages. In the present work a novel method of pattern mining is presented to solve the problem through FSTSOM. In this Paper, the proposed method is an improvement to the web log mining method and to the online navigational pattern forecasting. Here, Neural based approach i.e. Self Organizing Map (SOM) is used for clustering of sessions as a trend analysis. SOM depends on the clustering performance with the number of requests. In the proposed method, using the SOM algorithm for Frequent Sequential Traversal Pattern Mining called FSTSOM. In this method, first using SOM algorithm and getting some cluster of web-logs. Then loading that web-log cluster, which is nearly related to frequent pattern. After that applying Min-Max Weight of Page in Sequential Traversal Pattern. Finally, established good prediction with the number of data and the excellence of the results.