An iterative method for evaluating product quality and user reputation

With the development of the Internet, many online websites spring up, which provide a platform for users to purchase products. In order to evaluate the real quality of products, many of these websites introduce the rating system, where people can give discrete ratings to products to express whether they are satisfied or not about the transactions. Nevertheless, various forms of unfair attacks have been observed in the rating system. Therefore, how to evaluate the intrinsic quality of products becomes a significant yet difficult problem. There have been several works focusing on building a reputation system to solve this problem. In this paper, we consider that a user with a small rating fluctuation deserves more reputation. Based on this idea, we propose an efficient method to calculate the product quality in an iterative way. We use the standard deviation of the vector difference between the user's rating vector and the product's estimated quality vector to measure the fluctuation. The user reputation is inversely proportional to the product of the mean square between his rating vector and the products' estimated quality vector times the standard deviation. And then we redistribute the user's reputation according to the user's degree. The product quality is determined by the weighted average ratings. Our simulations demonstrate that our method can recognize the good quality product and give them high rankings. The proposed method has a better performance and robustness than the state of the art methods in both real and artificial networks.