Aimed at the defects of principal component cluster analysis, this paper puts forward two methods to improve it, and proves the new methods theoretically. The core distinctions between the new methods and similar methods lies in the new methods considered the influence of collinearity existed in cluster analysis, so that the new methods can reduce the error of classification model effectively. Besides that the new methods have abundant theoretical basis, and have general adaptation under similar complex classification problem. Empirical analysis shows that the improved principal component clustering analysis is valid.%针对主成分聚类分析的不足,本文提出了主成分聚类分析的两种改进方法,并对新方法进行了理论论证.新方法与同类方法的核心区别,在于考虑了聚类分析过程中变量共线性的影响,能够有效地降低分类模型的误差;理论基础充分,有着同类复杂分类问题下的普遍适应性.实证分析表明,改进的主成分聚类分析切实有效.
展开▼
