Abstract:
Bayes classifier, minimum distance classifier, nearest neighbour classifier and back propagation neural networks are widely used classifiers. In order to improve their performance, this paper introduces the idea of Box-Cox transformation. Box-Cox transformation, which can transform the data and make the distribution nearer normal distribution, is a simple but quite effective data processing technology. Experiment results show that, as the result of the introducation of Box-Cox transformation, the accurate rate of classifiers is improved remarkably.
Key words:
Box-Cox transformation,
Bayes classifier,
nearest neighbour classifier,
minimum distance classifier,
BP neural network
摘要: 贝叶斯分类器、最小距离分类器、近邻分类器和BP网络等是比较常用的分类器,为提高这些分类器的性能,引入了Box-Cox变换的思想。将Box-Cox变换用于数据正态化处理技术,并对常用分类器的性能进行改进。实验结果显示,通过引入Box-Cox变换,分类器的分类正确率有较大的提高。
关键词:
Box-Cox变换,
贝叶斯分类器,
近邻分类器,
最小距离分类器,
BP神经网络
CLC Number:
LI Jian-gang; WU Xiao-jun. Improvement of Classifier Performance Based on Box-Cox Transformation[J]. Computer Engineering, 2009, 35(23): 172-174.
李建刚;吴小俊. 基于Box-Cox变换的分类器性能改进[J]. 计算机工程, 2009, 35(23): 172-174.