Volume : VI, Issue : XII, December - 2017

LINEAR DISCRIMINANT CLASSIFIER BASED PEARSON CORRELATION FOR WEB TRAFFIC PATTERN MINING

Ulaganathan. N

Abstract :

 With the fast growing popularity of the WWW, Websites plays an important role to communicate knowledge and information to the users. The task of mining web traffic patterns is very difficult when the weblog database is enormous. Few research works has been designed for predicting the traffic web patterns to analyze the user’s behaviours. But, performance of existing traffic web patterns mining was not efficient. In order to overcome such limitations, MapReduce Pearson Correlation Fisher‘s Linear Discriminant Classifier (MPC–FLDC) technique is proposed. The MPC–FLDC technique efficiently extracts web traffic patterns from weblogs through pre–processing, classification and correlation analysis. The MPC–FLDC technique initially carried outs pre–processing in which MapReduce framework is used to group the web patterns according to different sessions. After preprocessing, MPC–FLDC technique applied Fisher‘s Linear Discriminant (FLD) Classifier to classify the web patterns at a different sessions as frequent or non–frequent based on hit ratio. This process resulting in improved classification accuracy of web patterns. Finally, MPC–FLDC technique used Pearson correlation analysis that evaluates the web patterns correlation between a different user sessions in order to efficiently predict the traffic web patterns with higher accuracy and minimum time. This process assists for MPC–FLDC technique to improve the traffic patterns prediction rate and reducing the prediction time. The performance of MPC–FLDC technique is evaluated with parameters such as classification accuracy, traffic patterns prediction accuracy, prediction time, and true positive rate with respect to different web patterns. The experimental result shows that MPC–FLDC technique is able to increases the traffic patterns prediction accuracy and also lessens the prediction time when compared to state–of–the–art–works.

Keywords :

Article: Download PDF   DOI : 10.36106/ijsr  

Cite This Article:

ULAGANATHAN.N, LINEAR DISCRIMINANT CLASSIFIER BASED PEARSON CORRELATION FOR WEB TRAFFIC PATTERN MINING, INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH : Volume-6 | Issue-12 | December-2017

Number of Downloads : 298

References :