An Adaptive Hoarding Technique (AHT) based on Naïve Bayes Classifier

Hoarding has become an attractive area for research in recent years. It suggests a good solution for one of the major mobile computing problems, which is the disconnection. Hoarding allows the continuity of computing in the absence of network connectivity, and hence, provides the network users with the illusion of a virtual network. The process of hoarding is to cache a subset of the user’s files needed in the next computing session on a local store prior to disconnection. However, hoarding is still a challenge as it requires predicting the user’s future behavior, then satisfying his future needs. The primary concern of this paper is to implement an efficient, transparent, and Adaptive Hoarding Technique (AHT) that operates without any user intervention. The proposed technique observes the user’s subject of interest, which was ignored by existing hoarding techniques, then predicts his future needs accordingly. AHT employs a modified version of naive bayes classifier to elect the highest quality files to be cached on the local hoard. Those files are the most relevant ones to the user’s subject of interest. This will significantly reduce the probability