Abstract: The paper presents a combination evaluation method based on parallel rank. Firstly, supposing theperformance of action is determined by utility of original decision matrix, the consistent results of a variety of single evaluation methods are treated as constraints of the utility function. Then actions are sorted by comprehensive utility and share parallel rank when they have equal utilities, which is the mean value of two goal-programing's solutions. Secondly, Weak-Kendall coefficient is defined to measure the consistency of two different ranks of actions, through which simulation experiments have been designed to analyse the robustness of combination evaluation methods when action number changes or original data are disturbed. At last, through a case from literature, using four single evaluation methods, four combination evaluation methods and Parallel-Rank method, the results show that Parallel-Rank method has better robustness. Future research directions have been given at the end of the paper according to the study on influence of the single evaluation methods selection to Parallel-Rank method.

Key words: combination evaluation, juxtaposition, Kendall coefficient, robustness

摘要： 本文提出一种允许并列排名的组合评价方法,利用多种单一评价方法对方案进行排序,提取多个排序结果中的一致优劣关系作为约束条件,并以原始决策矩阵为自变量构造效用函数,通过目标规划求解各方案的综合效用值,以综合效用值的大小确定各方案的排名,综合效用值相同的方案被认为没有差别,排名并列。定义Weak-Kendall系数描述评价方法的鲁棒性,从数据的随机扰动和方案数量的变化两个角度设计仿真实验,引入实例,并与离差最大化法、均值法、模糊Borda法等组合评价方法进行比较,结果表明并列排名组合评价方法在这以上两方面具有更强的鲁棒性。最后,通过研究单一评价方法的选择对并列排名法的影响,探讨方法的改进方向。

关键词: 组合评价, 并列排名, Kendall系数, 鲁棒性