A robust reputation iterative algorithm based on Z-statistics in a rating system with thorny objects

With the explosive growth of online alternative objects, the availability of a credible rating system could play a valuable role for users who are evaluating and/or choosing items. Due to the existence of spammers, many methods have been proposed that employ a reputation-based mechanism in the last few years. However, these methods only consider spammers' behaviours but neglect the effect of thorny objects, which are difficult to evaluate because of their intrinsic uncertainty and complexity with respect to the problem context, knowledge gaps, and various user criteria. To solve this problem, we propose a new reputation iterative algorithm based on Z-statistics (ZS), which eliminates the effect of thorny objects and strengthens the ability to deal with spammers. In this article, the proposed method makes effectiveness comparisons with other typical methods using a rating example and a simulated rating system, and presents additional effectiveness and complexity analyses of these results. Finally, the experimental results also demonstrate that the ZS algorithm outperforms other methods in simultaneously dealing with the effects of spammers and thorny objects.