对应分析降维效果的改进尝试

A Method to Improve the Effectiveness of Correspondence Analysis

目的 对应分析在给出低维投影结果时,有时会出现入选主因子累积贡献率值偏低的情形。本文尝试改进其分析结果。方法 将影响降维效果的个别变量点作为特例另行给出结论,不参与向量空间的特征描述。结果 入选主因子累积贡献率值增大,低维投影结果的可靠性提高。结论 资料的收敛特性制约着对应分析的降维效果,当影响收敛特性的向量点为个别变量点时,可考虑不将其引入对应分析运算,以改善低维空间对其余变量点的解释效果。

Objective The main factors in a low dimensional space may fail to explain much enough information in the primary data matrix.This article was supposed to provide a way to increase the cumulative contribution of the main factors.Methods Several variable points were picked out of the computation of the correspondence analysis,and were concluded from their vector velues.Results The cumulative contribution of the main factors was increased and the contribution from the map became more adequate.Conclusion The convergency of the primary data matrix influences the effect of the correspondence analysis.If there are only several vector points giving severe influences,they may be neglected from the description of the primary space,so as to make the low dimensional space explain other vector points effectively.