一种改进的基于密度聚类的支持向量机

Improved Support Vector Machine Based on Density Clustering

运用OPTICS算法能发现任意形状的聚类,且对输入参数不敏感的优势,提出一种基于OPTICS密度聚类的支持向量机算法,通过对原始数据进行预处理,利用可达图得到约简样本代替原始训练样本用支持向量机进行训练,降低了SVM训练所需的时间及空间复杂度.实验表明,该方法在保持分类精度的同时,大大缩短了训练时间,提高了分类效率.

Using the advantages of OPTICS algorithm which can discover clusters of arbitrary shape and not sensitive to the input parameters,an advanced Support Vector Machine algorithm based on the density clustering（ OPTICS） was proposed. The original data were pretreated through OPTICS algorithm which was utilized to get reachable distance graph. The original samples were replaced by the reduced samples to complete the training work for SVM,so that it reduced the training time and space complexity. Experimental results showed that this method can shorten the training time and improve the classification efficiency while maintaining the distribution of original samples and guaranteeing the classification precision.