CAJ | 학술논문

　　文本识别问题是模式分类中的一类重要的识别问题，也是较难处理的一类。该类问题中往往存在很多冗余属性，因此传统的分类方法对它的效果一般不好。本文针对文本识别问题，提出了一种基于核主成分分析的神经网络集成算法，该算法首先利用核主成分分析进行降维，合理的去除冗余属性，然后再利用神经网络集成算法进行分类学习。在文本分类数据集上的实验说明，本文算法可以有效地提高文本分类问题的分类性能。
　　문본식별문제시모식분류중적일류중요적식별문제，야시교난처리적일류。해류문제중왕왕존재흔다용여속성，인차전통적분류방법대타적효과일반불호。본문침대문본식별문제，제출료일충기우핵주성분분석적신경망락집성산법，해산법수선이용핵주성분분석진행강유，합리적거제용여속성，연후재이용신경망락집성산법진행분류학습。재문본분류수거집상적실험설명，본문산법가이유효지제고문본분류문제적분류성능。
Text recognition problem is an important class of recognition problems in pattern classification, and it is also more difficult to deal with. Since there is often a lot of redundant attributes for this kind of problem, so the effect of traditional classification methods is not very well. In this paper, for the problem of text recognition, a neural network ensemble algorithm based on kernel principal component analysis is proposed. The algorithm first use kernel principal component analysis to reduce the dimensionality, removing redundant attributes reasonable. Then use the neural network ensemble algorithm to classify. Experiments on text classification data sets illustrate the algorithm can effectively improve the classification performance of the text classification problem.