基于两阶段学习的专家群体评价信息集成方法研究

An Expert Group Evaluation Method Based on Two-Step Learning

专家群体偏好信息集成是综合评价定性指标测度与指标权重主观分配的基础。首先基于半可加离散分布统一描述的不同类型及精度专家偏好信息,建立第一阶段学习,统一其精度。然后重新界定群体偏好一致性概念,提出相适应检验方法,在此基础上建立第二阶段学习,调整非一致性的群体偏好至满足集成要求,并给出基于马尔科夫链的信息集成方法。上述两阶段学习通过模拟群体专家间的交互反馈机制,尽可能贴近偏好调整实际的同时无需专家参与,提高集成方法的可操作性及效率。最后,通过实例验证了基于两阶段学习的专家群体评价方法的可行性和有效性。

In the face of Individual preference information of diverse type and precision in the expert group evaluation,an expert group evaluation method based on two-step learning is proposed in this paper.Firstly,based on Individual preference information of diverse type and precision described by semi additive distribution,the firststep learning is established to unify its precision,and redefined the concept of consistency of group preference in addition with the adaptive consistency testing method.Secondly,the second-step learning is established to adjust non-consistency group preference to satisfy the integration demand,and the corresponding information integration method based on Markov chain is given.The two-step learning above tries hard to close to the actual preference adjustment without the participation of experts through simulating the interactive feedback mechanism during group experts,in order to enhance operability and efficiency of the integration method.Finally,an example shows the validity for the proposed method.