Reputation System User Classification Using a Hausdorff-Based Metric

In this paper we provide an automatic unsupervised recognition technique for Web community users, using a special nonlinear metric. First we propose a local trust metric to compute the ratings associated by an user to another user, in an explicit or implicit way. Our purpose is to help the users of the system to interact easily with users and resources they are interested in. User interest is computed using the proposed metric. On the basis of the obtained values, we propose a feature extraction approach for the reputation system users. The resulted feature vectors are clustered with an unsupervised classification algorithm using a special nonlinear metric.