Determining Similar Recommenders Using Improved Collaborative Filtering in MANETs

In a MANET environment, recommendation systems face a significant challenge whilst dealing with mobile nodes. Specifically, when a "node" is about to join a new cluster it may require some form of reference from a previously associated cluster. As such, research in this area has primarily focused on the selection of recommender nodes so that the overall consistency of the MANET system is maintained. In this study, an improved collaborative filtering mechanism has been exploited to address the selection of a group of suitable recommenders. First, a cluster formation algorithm has been used to group the set of recommenders based on their similarity measures with predictions computed independently for each cluster. Next, a threshold window is identified for selecting the best group of similar recommenders eliminating the lowest and highest trusted recommenders. Simulation results suggest that the proposed trust based similarity measures can greatly enhance the accuracy of node based trust management scheme.