基于科技项目评审专家异质性的共识测度方法研究

Research on consensus measurement method based on heterogeneity of scientific and technological project review experts

摘要:

共识度是共识识别过程的重要依据，直接对决策环境的共识进程产生影响，但一直以来共识度难以测度的重要原因之一是决策群体的异质性。在科技项目评审领域，专家共识度测算方法的缺失在某种程度上也阻碍了非共识项目的评审达成一致。因此，本文提出了一种基于科技项目评审数据的专家共识测度方法。此方法针对专家异质性难以测度问题，首先通过历史评审数据确定专家评审偏好，其次利用专家发文量、H指数等知识计算专家权威度，而后利用专家研究方向与评审项目研究内容得到主题覆盖度，三者结合得到专家权重；针对共识度的测算，利用专家权重，通过互反矩阵共识决策模型计算共识方案，再通过排名偏差重叠计算专家方案与共识方案的相似度得到共识度。本文提出的专家共识测度方法既结合了专家历史信息和当下信息，又考虑了专家权威度、研究方向等极易导致共识度产生偏差的异质性指标。最后，通过比较分析验证了方法的有效性与合理性。

Abstract:

The consensus degree is an important basis for the consensus identification process, which directly affects the consensus process of the decision-making environment, but one of the important reasons why the consensus degree has been difficult to measure is the heterogeneity of the decision-making group. In the field of scientific and technological project review, the lack of expert consensus measurement methods also hinders the review and consensus of non-consensus projects to some extent. Therefore, this paper proposes an expert consensus measurement method based on the review data of scientific and technological projects. Aiming at the problem that expert heterogeneity is difficult to measure, this method first determines the expert review preference through historical review data, secondly uses knowledge such as expert publication volume and H-index to calculate expert authority, and then uses expert research direction and review project research content to obtain topic coverage, and the three combine to obtain expert weight. For the calculation of consensus degree, the consensus scheme is calculated by using the expert weight, and the consensus scheme is calculated by the mutual inverse matrix consensus decision model, and then the similarity between the expert scheme and the consensus scheme is calculated by the overlap of ranking bias to obtain the consensus degree. The expert consensus measurement method proposed in this paper not only combines the historical information of experts and current information, but also considers heterogeneity indicators such as expert authority and research direction, which are easy to lead to consensus bias. Finally, the effectiveness and rationality of the method are verified by comparative analysis.