SepaRating: an approach to reputation computation based on rating separation in e-marketplace

This paper proposes SepaRating, a novel mechanism that separates a buyer's rating on a transaction into two kinds of scores: seller's score and item's score. SepaRating provides the reputation of sellers correctly based on the seller's score by repetitive separations, which helps potential buyers to find more reliable sellers. We verify the effectiveness of SepaRating via a series of experiments.