摘要： 为了提高模糊支持向量机在大数据集上的训练效率，提出一种基于两阶段聚类的模糊支持向量机算法。第1阶段为粗粒度聚类阶段，在每类训练样本上执行密度聚类算法，设置较大的邻域半径(给定邻域内最小点数)，保证可能成为支持向量的样本点都被选取；第2阶段为自适应聚类阶段，在粗选的数据集合上，执行自适应密度聚类算法，根据各个点距离分类面的远近，自适应决定该点的邻域半径(给定邻域内最小点数)。这样可有效地减少远离分类面的聚类边缘点的数量，同时在分类面附近保持较多的样本点，试验结果表明，基于两阶段聚类模糊支持向量机算法，相比以往的方法，不仅提高了模糊支持向量机的训练效率，同时保持了较好的分类效果。

关键词: 密度聚类, 支持向量, 模糊支持向量机

Abstract: To accelerate the training of Support Vector Machine(SVM), this paper proposes a new fuzzy support vector machine based on two stage clustering. After applying density clustering algorithm to the training samples of each class, the initial reduced training set is generated based on those clustering edge-samples. According to its distance towards separating plane, the neighbor radius of each sample is calculated. The new reduced training set is generated based on adaptive density clustering algorithm. Experimental results show that Fuzzy Support Vector Machine(FSVM) based on two stage clustering can get better classification results while reducing the training time greatly.

Key words: density cluster, support vector, Fuzzy Support Vector Machine(FSVM)

中图分类号: 

• TP301.6