Hierarchical Rough Fuzzy Self Organizing Map for Weblog Prediction

Abstract

Web servers play a vital role in conveying knowledge and information to end users. With rapid growth of WWW over past decades discovering hidden information about the usage pattern is critical towards determining effective strategies as well as to optimize server usage. Most of the available server analysis tools provide statistical data only without much useful information. Mining useful information becomes challenging task when user traffic data is huge and keeps on growing. In this work we propose hierarchical rough fuzzy self–organizing map (HRFSOM) to analyze useful information from the statistical data through weblog analyzer. We use cluster information generated by HRFSOM for data analysis and a variant of takagi sugeno fuzzy inference system (TSFIS) to predict daily and hourly traffic jam volumes. The experiments are performed using web user access sample patterns available at Yandex Personalized Web Search Challenge where statistical weblog data is generated by AWStats web access log file analyzer. The proposed classifier has superior clustering accuracy compared to other classifiers. The experimental results demonstrate the efficiency of proposed approach.