改进的并行SVM回归算法

西安科技大学 计算机科学与技术学院,陕西 西安 710054

支持向量机; K-means聚类; 并行计算; 顺次最小优化算法

Impoved parallel SVM regression algorithm
SHE Xiang-yang,CUI Wen-qiang

(College of Computer Science and Engineering,Xi'an University of Science and Technology,Xi'an 710054,China)

support vector machine; K-means clustering; parallel computing; sequential minimal optimization algorithm

DOI: 10.13800/j.cnki.xakjdxxb.2017.0223

备注

针对目前SVM回归算法在大样本情况下,学习效率低、精度不高的问题,提出了基于K-means聚类的并行SVM回归算法。在Hadoop框架中,先对训练样本行进并行聚类,然后针对聚类后的不同簇,构造相应的SVM回归模型,使用顺次最小优化算法求解各模型参数。预测时,选择与待预测样本距离最近簇的对应SVM回归模型进行预测。实验验证了文中算法的可行性和有效性。

Aiming at the low efficiency and poor accuracy problem of serial traditional SVM regression algorithm in large sample cases,the improved parallel SVM regression algorithm based on K-means clustering is proposed.In the framework of Hadoop,the training sample is clustered parallelly,then the SVM regression model is constructed for every cluster,the parameters of the SVM regression model are solved by sequential minimal optimization algorithm.The cluster regression model of the nearest cluster to test sample is selected to predict.Algorithm testing show:!![Impoved parallel SVM regression algorithm] is feasible andeffective.