Web Log Mining for Improvement of Caching Performance

The objective of this study is to build a model of cache replacement policy for improvement of web caching performance. The integration approach of cluster analysis and classification are used to create a classifier for predicting the cache life time. The data set was collected from the cache of the National Institute Development Administration's proxy servers. The data collection period was from October 2009 to March 2010. There are four main tasks in this study. First, the access log from proxy servers were collected and preprocessing tasks were performed. Second, the access log data were partitioned into clusters based on users' access patterns. Third, classifier models of the cache replacement policy were built and their accuracies were compared. Finally, the efficiency of the selected classifier was compared with other cache replacement algorithms. Results show that overall classification accuracy of the model is satisfactory and the model is efficient and very good in performance.