Rating can be obligatory for many tasks, such as film recommendation, hotel rating, and product evaluation. Aggregating ratings given by numerous raters is a necessary and effective way to obtain comprehensive evaluation of the objects. While the awareness of potential distortion for some of the targeted objects, has attracted substantial attention of researchers and motivated the designing of the robust rating aggregation method to overcome the impact of disturbance from ignorant/malicious raters in practice. In this paper, we focus on rating aggregation with collusive disturbance, which is hard to be eliminated and invalidate traditional rating aggregation methods. Therefore, we will introduce the idea of detecting collusive group into rating aggregation to develop a new method, called robust rating aggregation method based on rater group trustworthiness (RGT), which obtains four main modules: Graph Mapping, Rater Group Detection, Group Trustworthiness Calculating, and Rating Aggregation. Experimental results and analyses demonstrate that our method is more robust to collusive disturbance than other traditional methods.

许多任务都必须进行评级，例如电影推荐、酒店评级和产品评估。汇总众多评价者的评分是获得对评价对象综合评价的必要且有效的方法。虽然对某些目标对象潜在失真的认识引起了研究人员的广泛关注，并促使设计鲁棒的评分聚合方法，以克服实践中无知/恶意评分者干扰的影响。在本文中，我们关注的是具有共谋干扰的评级聚合，这种干扰很难消除，并且使传统的评级聚合方法失效。因此，我们将检测共谋群体的思想引入评分聚合中，开发一种新的方法，称为基于评分者群体可信度（RGT）的鲁棒评分聚合方法，该方法获得四个主要模块：图映射、评分者群体检测、群体可信度计算，以及评级聚合。实验结果和分析表明，我们的方法比其他传统方法更能抵抗串通干扰。