Online Reputation Scoring from Historical Feedbacks Based on Partially-Ordered Similarity.

Abstract: Online reputation evaluation has been implemented in many practices in different ways. Based on the evaluation of several reputation mechanisms, this paper proposes a reputation evaluation for scoring and monitoring the seller''s behavior. Second-order atomic patterns are defined, based on which we propose an algorithm named Partially- Ordered Pattern Scoring (POPS). A solution for reputation monitoring is discussed. This mechanism helps to prevent several cheating strategies. At the end of the paper, comparative experiments on accuracy with Double Exponential Smoothing and on stability with different lengths of cases have shown its advantages. [Copyright &y& Elsevier]