摘要
传统的支持向量机分类算法在优化过程中对所有支持向量都进行优化,增加了计算量,降低了训练效率.针对上述缺点,在分析样本模糊隶属关系的基础上,采用改进的K近邻算法为已知样本分配隶属度,根据训练样本的隶属关系,剔除非支持向量,减少训练样本,并将其用于中文网页的分类中,得到了较好的分类效果.仿真实验结果表明,改进后的方法不仅相对简单,而且在保证分类器性能的情况下,能有效地减少支持向量机的训练样本数,从而提高支持向量机的训练和测试速度.
Traditional Support Vector Machine （SVM） algorithms for classification optimize all support vectors in the optimiza- tion process. They increase the amount of computation and reduce the efficiency of training. Aiming at the above shortcomings, it a- dopts an improved K-nearest neighbor algorithm to assign affiliation to known samples on the basis of analysis sample fuzzy affilia- tion, which eliminates the non - support vector and reduces the training samples. This method was used for classification of Chinese web pages and obtained better classification results. The simulation results show that the improved approach is not only simpleness, but also can effectively reduce the number of training samples of SVM and enhance the training and testing speed of SVM in the case of ensuring classifier performance.
出处
《渭南师范学院学报》
2012年第2期83-86,共4页
Journal of Weinan Normal University
基金
陕西省教育厅科研资助项目(2010JK095)
关键词
K近邻算法
支持向量
分类
K- nearest neighbor algorithm
support vector
classification
