中医舌色苔色识别中多分类支撑向量机算法比较

Comparisons for Multi-Class Support Vector Machine on Tongue Color Recognition of Traditional Chinese Medicine

为了了解在不同样本数量情况下具有何种特点的算法比较适合舌色苔色的分类与识别,本文采用一对一SVM和它的改进算法IMP-BSVM、DAG-BSVM,以及M-ary分类SVM算法对舌色、苔色分类进行了比较。结果表明,当训练样本数较少时,DAG-BSVM能有较高的正确率和较快的训练和识别速度,因此选用DAG-BSVM;当训练样本数较大时,IMP-BSVM能得到最少的支撑向量和较高的正确率以及较快的识别速度,此时则选用IMP-BSVM。

In this paper, in order to realize which kind of algorithm is suitable for tongue color classification and recognition when the number of samples is small or large, we detailedly introduce one-against-one SVM and its improved algorithm IMP-BSVM, DAG-BSVM, and M-ary classification methods on basic thinking, implements, characteristics and existent problems based on tongue color classification and recognition of traditional Chinese medicine. The results indicate when the number of training samples is less, DAG-BSVM has higher correct rate, training speed and testing speed, So the metheod of DAG-BSVM is suitable. When the number of training samples is great, IMP-BSVM produces the smallest support vector and has higher correct rate and testing speed. So IMP-BSVM is suitable in the case.