中医舌象样本分类中加权SVM的应用研究

Weighted Support Vector Machine for Classification of Tongue Images

中文关键词:  加权SVM  舌象识别  分类器  舌诊

英文关键词:weighted SVM  tongue images recognition  classifier  tongue inspection

中文摘要:

      SVM在机器学习领域中是一种通用有效的学习方法,在中医舌象分类中取得了一定的效果。但是由于各类舌象样本的数量不均衡,标准SVM方法对于这一问题的解决尚不尽人意。针对这一问题,本研究将加权SVM方法应用于大量舌色苔色样本的分类中,根据各类样本的相对重要性,加大或减小其惩罚项,在保证非重要样本的识别正确率可以接受的情况下,尽量提高重要样本的识别正确率。实验结果表明,该方法在不均衡舌象样本的分类中能取得较好的效果。

英文摘要:

      As an effective learning method in the machine learning，support vector machine has been used in the classification and recognition of tongue images.Although it gained achievements to a certain extent，the result was not satisfactory due to unbalanced size in tongue samples belonging to different categories.To solve this problem，an alternative method called weighted support vector machine is introduced.Its basic idea is to assign different weights on penalty term to the samples from two different classes respectively according to the relative importance.Thus the recognition accuracy of the more important data is improved while the accuracy rate of the less important samples is still acceptable.Experimental results show that the method works well in the classification of unbalanced tongue images.