面向误差最小化的在线服务信誉度量

由于每个在线服务可以通过其自身的真实质量被客观比较,存在潜在的真相服务排序。为了使用户进行服务选择时有真实客观的在线服务信誉排序作为参考,服务信誉应当尽可能地接近真相服务排序。提出一种面向误差最小化的在线服务信誉度量方法。该方法将用户对服务的偏好排序视为对真实服务排序的带噪估计,利用Kendall tau距离指标来衡量服务排序与真相排序之间的误差,通过设定真相与用户对服务的偏好排序集合之间的平均误差上限找出可能的真相服务排序,寻找与可能的真相服务排序集合之间平均误差最小的服务排序作为服务信誉。由于所有的服务排序都有可能为真相排序,造成了该方法的计算困难,利用分支切割法对该方法进行优化求解。以真实数据集和模拟数据集为基础,通过实验验证了该方法在保证运行效率的同时得到与真相误差更小的信誉度量结果。

Online service reputation measurement for error minimization

Since each online service can be objectively compared by its own real quality, there is a potential truth ranking of services. In order to provide users with the most authentic and objective online service reputation ranking as a reference for choosing services, service reputation should be as close as possible to the true service ranking. In this paper, an online service reputation measurement method for error minimization was proposed and it regarded user preference ranking as a noisy estimation of real service ranking. Firstly, Kendall tau distance was used to measure the error between service ranking and truth ranking. Then, the possible ranking of truth services was found by setting the upper limit of the average error between the truth and the user''s preference ranking set. Finally, the service ranking with minimum average error between itself and the possible sets of service ranking was found as the service reputation. Because all the service ranking could be the truth ranking, causing the computational difficulty of this method, the branch-and-cut algorithm was used to optimize the solution. Based on the real and simulated data sets, experiments were carried out and the result showed that reputation measurement results could be obtained with less error between it and the truth while ensuring the operation efficiency.