Building and managing reputation in the environment of Chinese e-commerce: a case study on Taobao

We propose a lightweight reputation model R-Rep, for resisting manipulative behavior in Reputation Systems. We present a manipulative behavior detection system CSI to detect the customers who provide manipulative ratings, and the vendors who intend to increase their reputation value in a strategic manner. Via analysing motivation of manipulative behavior, we specify features for identifying suspicious customers using clustering algorithm. Utilizing the inherent relationship between suspicious customers and suspicious vendors, the first set of suspicious vendors is identified by CSI. Meanwhile, using different pieces of information, which refer to non-anonymous ratings and anonymous ratings, the second and the third sets of suspicious vendors are detected by CSI. We designed two universal approaches RVA and BVA to compare different reputation models with regard to resisting manipulative behavior. The comparison approaches are applied to a set of suspicious vendors identified by CSI. Results show that, R-Rep outperforms two existing models, the reputation model employed by Taobao (the largest e-commerce site in China) and a Bayesian System. The two comparing approaches RVA and BVA have inherent consistency.