基于群体偏好的交易评价可信度
Transaction rating credibility based on user group preference

利用交易评价数据对商品和卖家进行信用评价以供用户参考成为电子商务在线交易平台最基本的服务。然而,目前的信用评价方法很少考虑不同用户间的评价偏好差异,将所有用户的评价同等看待,导致蓄意刷分或恶意差评等信用造假问题屡禁不止。该文提出了一种基于群体偏好的交易评价可信度确立方法。首先采用K-means聚类算法将用户分为3类用户群,通过实证数据分析验证了用户群间明显的评价偏好差异,然后利用评价偏好特征确立每类用户不同类型交易评价的可信度,并提出了动态的交易评价可信度更新策略。该方法能够有效地限制信用造假行为。
Abstract：Transaction and rating data can be used to evaluate credit as a key underlying service provided by online transaction platforms. However, credit evaluation methods pay little attention to user rating preferences, so rating manipulations can expand arbitrarily. A method is given here to determine transaction rating credibility based on group preferences. This method uses the K-means clustering algorithm to divide all the users into three groups and analyzes and validates each group's rating preferences. Then, the algorithm provides three steps to determine the credibility of ratings on different levels for different user groups based on these preferences. This paper also provides a strategy to dynamically update the group division and credibility to effectively restrict credit manipulation.