中国区域创新能力静态分析——基于自适应赋权主成分聚类模型  

Static Analysis of China's Regional Innovation Capability——Based on Adaptive Weighting Principal Component Cluster Model

作　　者：朱建平[1,2] 王德青[1,2] 方匡南[1,2] 

机构地区：[1]厦门大学经济学院统计系,福建厦门361000 [2]厦门大学数据挖掘研究中心,福建厦门361000

出　　处：《数理统计与管理》2013年第5期761-768,共8页Journal of Applied Statistics and Management

基　　金：国家社会科学基金项目(编号:11BTJ001);国家自然科学基金青年项目(编号:710201139);全国统计科学研究计划重大项目(编号:2012LD001)

摘　　要：本文在对经典聚类模型和现有改进聚类模型优点与不足剖析的基础上,通过定义客观加权主成分距离为分类统计量,提出了一种自适应赋权的主成分聚类模型。与现有同类方法相比,新模型克服了指标之间的高度共线性,能够对指标重要性的客观差异进行自适应赋权,每一步都有充分的理论保证其必要性、合理性.应用加权主成分聚类对中国区域创新能力进行集团划分,分类结果的可解释性明显提高,统计检验效果显著,所得的结论对了解和推动中国区域创新能力发展具有借鉴意义。Based on summation of advantages and defects of classical cluster models and existed improved cluster models, this paper puts forward an adaptive weighting principal component cluster model by defining objective weighted principal component distance as classification statistic. Compared with existed similar methods, the new model not only solves high coiinearity among variables, but also weights adaptively according to their objective importance, so the new model deserves sufficient theoretical basis for its necessity and rationality at every stages. At last this paper applies weighted principal component cluster model on the division of China＇s regional innovation capability, and gets highly improved explana- tion results as well as significant statistical test. Meanwhile the conclusion provides significant reference to understand and promote China＇s regional innovation capability.

关 键 词：分类 自适应赋权 加权主成分聚类 区域创新能力 

分 类 号：C812[社会学—统计学] O212[理学—概率论与数理统计]