应用PCA和RBF网络评价建设项目动态联盟候选伙伴的绩效  

Evaluating Candidate Partners' Performance of Dynamic Alliance of Construction Project by Means of PCA and RBF Network

作　　者：刘雷[1] 李南[1] 

机构地区：[1]南京航空航天大学经济与管理学院,江苏南京210016

出　　处：《系统工程》2008年第4期11-17,共7页Systems Engineering

基　　金：国家自然科学基金资助项目(70473037);国家软科学研究计划资助项目(2006GXQ3B184)

摘　　要：基于建设项目动态联盟候选伙伴绩效的内涵分析,确定了决定伙伴绩效的因素,构建了伙伴绩效评价的指标体系。通过主成分分析将众多指标进行综合,消除样本间的信息重叠,降低RBF网络的输入维数。针对伙伴绩效系统的非线性特征,采用RBF网络高度非线性映射能力,对某建设项目动态联盟的候选伙伴绩效进行了评价。评价结果表明PCA与RBF网络相结合的方法比单纯的RBF网络方法具有较高的精确度和较好的拟合效果。Based on the analysis of the meaning of candidate partners＇ performance of dynamic alliance of construction project, we have ascertained the determinants of partners＇ performance, and established the evaluation criteria system for partners＇ performance. Through principal component analysis, we have synthesized numerous criteria, eliminated information overlapping of the ,sample, and reduced the input dimension of RBF network. According to the nonlinear feature of partners＇ performance system, we have used RBF network altitudinal nonlinear map to evaluate candidate partners＇ performance for a dynamic alliance of construction project. The results show that the conjoint method — PCA and RBF network is more precise and fits better than the single RBF network method.

关 键 词：候选伙伴绩效评价 建设项目动态联盟 主成分分析法 RBF神经网络 

分 类 号：F284[经济管理—国民经济] F272