PROBLEMS DISCUSSION ON THE APPLICATION OF PCA IN COMPREHENSIVE EVALUATION

摘要

More and more people like to use the method of principle component analysis (PCA) in comprehensive evaluations recently. Many of them use the contribution ratios of variances as the weights to make linear combination. Someone has pointed out that it is not reasonable because of the multiple correlations among the indexes and put forward a few resolving methods, such as only using the first principle component or correcting with the subjective weights of importance. This paper finds out the above-mentioned methods are not efficient to satisfy the requirement of comprehensive evaluations with these methods, and the important place should be given to the notions of evaluation purpose, and the aggregate index plays the key role. The completeness of index system could be improved by using the method of PCA, but the contribution ratios of variances can not be used as the weights. Further it is not appropriate to use the conception of information quantity in the method. Finally, the mentioned principles are illustrated by a figure example.