基于混合模糊聚类的项目评审专家匹配方法  

Matching method for project review experts based on hybrid fuzzy clustering

作　　者：高亚琦 游子毅[1] 杨乘[1] 李思瑶 GAO Ya-qi;YOU Zi-yi;YANG Cheng;LI Si-yao(School of Physics and Electronic Science,Guizhou Normal University,Guiyang 550025,China;Software R&D Department,Guizhou Software Engineering Research Center Co.Ltd,Guiyang 550000,China)

机构地区：[1]贵州师范大学物理与电子科学学院,贵州贵阳550025 [2]贵州省软件工程研究中心有限责任公司软件研发部,贵州贵阳550000

出　　处：《计算机工程与设计》2025年第4期1022-1029,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(62062025);贵州省科技支撑计划基金项目(一般项目)(黔科合支撑[2023]一般333)。

摘　　要：针对学科交叉领域项目与专家间的相似性特征挖掘及精确匹配问题,提出一种基于混合模糊聚类的评审专家智能匹配方法。基于专家信息关联库筛选出与项目主题领域高相似度范围内的专家;在此范围内,采用改进的密度峰值聚类的中心选择方案确定专家-项目主题向量集的初始簇中心,并提出基于加权模糊相似度的K均值迭代完成簇区成员的优化;依据先回避、再动态抽取的策略完成评审专家匹配和推荐。在项目评审真实数据集的实验结果表明,所提方法对于大多数项目表现出较高的匹配精度和稳定性,符合交叉领域项目专家匹配的多样性需求。To address the problem of deeply mining the hidden similar features between interdisciplinary projects and experts,an intelligent matching method for project review experts based on hybrid fuzzy clustering algorithm was proposed.A range of experts with high similarity to the project’s professional field was obtained from the expert information database.Based on this range,an improved center selection scheme of density peak clustering was adopted to determine the initial center points,and a weighted fuzzy similarity measure based K-Mmeans iteration was proposed to optimize the members of each cluster.The recommendation of review experts was completed from the clustering results according to a strategy that first avoiding and then dyna-mically extracting.The experimental results on the real dataset of project reviews substantiate that,the proposed method exhi-bits high matching accuracy and stability for most projects,which meets the diverse needs of expert matching in interdisciplinary projects.

关 键 词：学科交叉领域 评审专家推荐 混合模糊聚类 密度峰值 语义匹配 相似性挖掘 回避策略 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]