| 基于信息贡献率的评价指标筛选与赋权方法 |
| |
| 引用本文： | 陈洪海,王慧,隋新.基于信息贡献率的评价指标筛选与赋权方法[J].科研管理,2020,41(8):240-247. |
| |
| 作者姓名： | 陈洪海 王慧 隋新 |
| |
| 作者单位： | 南京财经大学 金融学院，江苏 南京210023 |
| |
| 基金项目： | 国家自然科学基金;江苏省高等学校自然科学研究项目;教育部人文社会科学研究项目;国家社会科学基金 |
| |
| 摘 要： | 针对现有因子分析及相关分析两种方法在指标筛选与赋权方面有待完善之处，提出一种基于信息贡献率的评价指标筛选与赋权方法。该方法首先利用因子方差贡献率和因子载荷构建评价指标的信息贡献率，以此反映该指标解释原始指标集全部信息的比例，然后在现有相关分析法中引入病态指数测度指标集的信息重叠水平，进而实现评价指标的系统筛选与赋权。本研究认为评价指标的信息贡献率越大，其解释原始指标集信息的比例越大，因此该指标越应予以保留，其权重亦应越大。此外，研究亦发现因子分析法易误删信息解释能力强的指标，相关分析法筛选指标易不充分，而本文方法解决了这些问题。
|
| 关 键 词： | 评价指标 指标筛选 指标赋权 信息贡献率 信息重叠 |
| 收稿时间： | 2017-08-28 |
| 修稿时间： | 2018-07-16 |
Evaluation indicator screening and weighting method based on information contribution ratio |
| |
| Chen Honghai,Wang Hui,Sui Xin.Evaluation indicator screening and weighting method based on information contribution ratio[J].Science Research Management,2020,41(8):240-247. |
| |
| Authors: | Chen Honghai Wang Hui Sui Xin |
| |
| Institution: | School of Finance, Nanjing University of Finance and Economics, Nanjing 210023, Jiangsu, China |
| |
| Abstract: | The evaluation indicator system is the basis of comprehensive evaluation. Whether the selection of evaluation indicators is scientific or not is related to whether the construction of evaluation indicator system is reasonable or not, and it is meaningless to choose the best evaluation method if selection of evaluation indicators is unreasonable. Therefore, how to select evaluation indicators scientifically has always been concerned by researchers.
Factor analysis and correlation analysis are two kinds of evaluation indicators screening methods which are widely used at present. The purpose of factor analysis is to eliminate the indicators with weak impact on the evaluation results, while the purpose of correlation analysis is to reduce the level of information overlap between the evaluation indicators, so as to reduce distortion of information overlap on comprehensive evaluation results. However, the existing factor analysis indicators screening method is only based on a certain factor load to represent the level of information in the original indicator set interpreted by the indicator, which omits the important role of other factor loads of an indicator in the interpretation of original indicators set information, and cannot comprehensively and accurately represent the level of an indicator to interpret original indicator set information. At the same time, although existing correlation analysis method can reduce the information overlap of indicator set to a certain extent, it is unable to judge whether the overall information overlap between the remaining evaluation indicators is low, and whether it is necessary to further select indicators, so it is very easy to cause the excessive or inadequate of indicators selection.
In views of the shortcomings of the existing factors analysis and correlation analysis methods, a new method is proposed to screen and weight the evaluation indicator based on information contribution ratio. The information contribution ratio is constructed by factor variance contribution rate and factor loading. The ill-conditioned indicator is introduced in the existing correlation analysis to measure the information overlap level of the indicator set. On this basis, systematic screening and weighting of the evaluation indicators are realized. Finally, this paper takes 500 small business credit business data of a commercial bank in China as an example, and compares the effect of indicators selection with the existing factor analysis and correlation analysis from two aspects of information interpretation ability and overall information overlapping reduction level.
This paper suggests that the greater the information contribution ratio of the evaluation indicator, the larger the proportion of the explained information of the original indicator set; so the more the indicator should be retained, the greater weight of the indicator should be. It is found that compared with the existing factor analysis method, the information interpretation ability of the retained evaluation indicators by this method proposed in this paper is stronger. At the same time, compared with the existing correlation analysis method, this method can effectively reduce the overall information overlap level of the indicator set. In addition, both of the problems are found in the research and solved by the proposed method that the factor analysis method is easy to delete the indicators with strong ability to information explanation and the correlation analysis method is insufficient in indicators screening. |
| |
| Keywords: | evaluation indicator indicator screening indicator weighting information contribution ratio information overlap |
| 本文献已被 万方数据 等数据库收录！ |
| 点击此处可从《科研管理》浏览原始摘要信息 |
| 点击此处可从《科研管理》下载免费的PDF全文 |
|
