有序群决策中基于SVDD的专家排序方法研究

Study on Expert Ranking Method Based on SVDD in Ordered Group Decision

针对有序群决策问题,本文利用支持向量域描述给出一种专家权重的确定方法。首先把每位专家对所有方案给出的估计值视为高维空间特征向量,其次在使用该方法学习时采用"留一法"的思想,即分别留下一个专家的特征向量,把其他所有专家的特征向量进行描述,分别得到一个最小包球。然后计算每个专家的特征向量到其他所有专家的特征向量所得最小包球的球心的距离,最后把每个距离的倒数作为衡量该专家体对群体意见的贡献,从而计算出每位专家的评判权重,并依据权重的大小进行排序,算例显示该方法的有效性和应用性。

A method for determining the weights of experts in an ordered group decision-making was given by using the support vector data description.Firstly,the estimates given by each expert for all the schemes were regarded as a high-dimensional space feature vector.Secondly,when using the method to learn,the idea of"leave one method"is adopted,that is,to leave an expert's feature vector separately,describe the feature vectors of all other experts,and get a minimum package ball respectively.Then the distance from the feature vector of each expert to the center of the smallest envelope sphere obtained by the feature vectors of all other experts was used as the basis.Finally,the reciprocal of each distance is used to measure the contribution of the expert body to the opinions of the group,thereby calculating the evaluation weight of each expert,and sorting according to the weight.A numerical example shows the effectiveness of this method and the applicability of the method.