论文部分内容阅读
一套好的指标应该能够以较少的信息高效反应更多的内容,如果对指标不加选择地应用,很容易造成回归中的共线性、智能算法中的信息空间过大等问题。本文从降低指标间的相关性入手,通过建立一套新的方法减少信息冗余。这套方法以因子分子为主,充分利用旋转后因子载荷矩阵的信息,并结合聚类分组、相关分析等。以50家上市公司2015年的数据为样本进行实证分析,最终结果显示这一方法能有效降低数据间的相关程度,达到了精简指标的目的。
A set of good indicators should be able to respond more efficiently with less information. If the indicators are not used selectively, it is easy to cause collinearity in regression and too much information space in intelligent algorithms. This article starts with reducing the correlation between indicators and reduces information redundancy by establishing a new set of methods. This method is based on factor molecules, making full use of the information of the rotated factor loading matrix, combined with clustering grouping and correlation analysis. The empirical analysis of the data from 50 listed companies in 2015 shows that this method can effectively reduce the correlation between data and achieve the purpose of streamlining indicators.