Improving the Performance of a Proxy Cache Using Expectation Maximization with Naive Bayes Classifier

Abstract

AbstractThe Expectation Maximization Naive Bayes classifier has been a centre of attention in the area of Web data classification. In this work, we seek to improve the operation of the traditional Web proxy cache replacement policies such as LRU and GDSF by assimilating semi supervised machine learning technique for raising the operation of the Web proxy cache. Web proxy caching is utilized to improve performance of the Proxy server. Web proxy cache reduces both network traffic and response period. In the beginning section of this paper, semi supervised learning method as an Expectation Maximization Naive Bayes classifier (EM-NB) to train from proxy log files and predict the class of web objects to be revisited or not. In the second part, an Expectation Multinomial Naïve Bayes classifier (EM-NB) is incorporated with traditional Web proxy caching policies to form novel caching approaches known as EMNB-LRU and EMNB-GDSF. These proposed EMNB-LRU and EMNB-GDSF significantly improve the performances of LRU and GDSF respectively.KeywordsWeb cachingProxy serverCache replacementClassificationExpectation Maximization Naive Bayes classifier