一种基于支持向量机的模糊分类器

Fuzzy Classifier Based on Support Vector Machine

提出了一种基于支持向量机学习的模糊分类器(FCBSVM)。介绍了FCBSVM的基本思想及其结构,分析了隶属函数参数和惩罚参数C对分类规则的产生以及分类性能的影响,并提出了参数确定方法。构建这种分类器时,先选用适当的隶属函数,构造核函数。然后,以训练模式作为中心,进行模糊划分,对每个模糊划分建立一条模糊IF-THEN分类规则。最后,利用支持向量机学习方法,求出支持向量和规则的参数。这种分类器将支持向量机和模糊集合理论的优点结合起来,实现了模糊划分和模糊分类规则的自动产生。用双螺旋线数据和典型的数据集对分类器的性能进行了实验评测,验证了分类器的有效性。

A fuzzy classifier based on Support Vector Machine （FCBSVM） was proposed. The basic idea and the structure of this classifier were introduced. The effects of the membership functions parameters and the penalty parameters for the classification rule and the classifier performance were analyzed. The method of selecting rules was described. For the process of the proposed classifier constructed, firstly the kernel function was constructed by selecting suitable membership function, then a fuzzy partition was built around each training pattern and a fuzzy IF-THEN classification rule was defined for each fuzzy partition, finally the support vectors and the parameters for rules were got by SVM learning method. Since this classifier has the virtues of fuzzy set and Support Vector Machine, the fuz~ partition and the classification rules can be produced automatically. Experiments with two-spiral data and typical data sets evaluate the performances of the proposed classifier.