Web pre-fetching schemes using Machine Learning for Mobile Cloud Computing

Abstract

Pre-fetching is one of the technologies used in reducing latency on network traffic on the Internet. We propose this technology to utilise Mobile Cloud Computing (MCC) environment to handle latency issues in context of data management. However, overaggressive use of the pre-fetching technique causes overhead and slows down the system performance since pre-fetching the wrong objects data wastes the storage capacity of a mobile device. Many studies have been using Machine Learning (ML) to solve such issues. However, in MCC environment, the pre-fetching using ML is not widely used. Therefore, this research aims to implement ML techniques to classify the web objects that require decision rules. These decision rules are generated using few ML algorithms such as J48, Random Tree (RT), Naive Bayes (NB) and Rough Set (RS).These rules represent the characteristics of the input data accordingly. The experimental results reveal that J48 performs well in classifying the web objects for all three different datasets with testing accuracy of 95.49%, 98.28% and 97.9% for the UTM blog data, IRCache, and Proxy Cloud Computing (CC) datasets respectively. It shows that J48 algorithm is capable to handle better cloud data management with good recommendation to users with or without the cloud storage.