基于多元对应分析的KNN分类器组合

COMBINATION OF KNN CLASSIFIERS BASED ON MULTIPLE CORRESPONDENCE ANALYSIS

本文提出一种基于多元相应分析的ＫＮＮ分类器组合方法（ＭＣＡＫＮＮ），并以手写体识别为例，用ＫＮＮ 分类器在同一样本集合得到的不同特征集上进行分类，再通过多元对应分析对这些分类器的结果进行组合，以得到最终的分类结果．实验结果表明，此种分类器组合方法能显著减少分类错误率．

This paper presents a KNN classifier combination method based on multiple correspondence analysis (MCA KNN). This combination method is applied to written character recognition. Four kinds of features are extracted from same sample set and four result sets are obtained from these feature sets through KNN classifier. Through MCA KNN, the four result sets are combined to get the final result. The experimental results in this paper demonstrate MCA KNN's capability to reduce classifying error rate.