计算机与现代化Issue(4)：22-25,4.DOI:10.3969/j.issn.1006-2475.2018.04.005
基于改进k-近邻的直推式支持向量机学习算法
TSVM Learning Algorithm Based on Improved K-nearest Neighbor
摘要
Abstract
TSVM traverses all unlabeled samples,which makes it time-consuming.This paper proposes k2TSVM,a new TSVM algorithm based on improved k-nearest neighbor.Unlabeled samples are first clustered with k-means,then the k-nearest neighbors of each cluster centroid are found.Based on the distribution of positive and negative data, certain samples are deleted,and the remaining data are fed into TSVM.k2TSVM does not need to traverse all unlabeled data as TSVM does,so it reduces the training set size and improves training speed.Experiments show that k2TSVM achieves better classification accuracy in comparison to sim-ilar revised TSVM algorithms,while reducing running time.关键词
支持向量机/直推式学习/k-近邻法/k-均值聚类/无标签样本Key words
support vector machine/transductive inference/k-nearest neighbor/k-means clustering/unlabeled samples分类
信息技术与安全科学引用本文复制引用
李煜,冯翱,邹书蓉..基于改进k-近邻的直推式支持向量机学习算法[J].计算机与现代化,2018,(4):22-25,4.基金项目
四川省科技厅重点研发项目(2017GZ0331) （2017GZ0331）
