摘要
模糊支持向量机(FSVM)具有很好的抗噪声能力,受到了很多专家的重视。然而模糊支持向量机算法的时间复杂度通常较高。针对这一不足,本文提出了一种基于核聚类的模糊支持向量机算法。首先根据核聚类算法对每一类原始样本进行聚类,然后对每一簇求样本中心,用样本中心作为新的样本点替换该类别的原始样本。最后本文算法利用中心距离型计算新样本的模糊权重,并利用模糊支持向量机算法进行求解。实验充分验证了本文算法相对于传统模糊支持向量机方法具有更快的分类速度。
As fuzzy support vector machine （FSVM） is with good noise immunity, it has received the attention of many experts. However, the time complexity of FSVM is usually higher. For this shortage, this paper presents a FSVM algorithm based on kernel clustering. Firstly, it clusters the original samples of each type, then computes the mean of each cluster and constructs the new training set. Finally, it uses the center distance type method to compute the fuzzy weight, and performs FSVM training. The experiment fully validated this algorithm has faster classification efficiency than traditional FSVM.
出处
《科技通报》
北大核心
2013年第10期133-135,共3页
Bulletin of Science and Technology
关键词
模糊支持向量机
抗噪声
核聚类
中心距离型
fuzzy support vector machine
noise immunity
kernel clustering
center distance type
