摘要
经典支持向量机算法具有较高的时空复杂度,这导致其很难广泛被应用.为此,该文基于支持向量分布的先验知识,把KNN算法和10折交叉验证方法结合起来,提出了一个支持向量预选取算法.该算法从原始样本集中选取k个邻近样本,并计算出这k个邻近样本中异类样本所占比例,如果该比例超过了所给定的阈值q,就选择这些异类样本作为预取的支持向量.在此过程中,采用10折交叉验证法来确定k与q的最佳值.两组仿真实验表明所提算法选择出的支持向量而形成的分类器分类准确率更高而且耗时还较少.
According to shortcoming of higher time and space complexity of classical support vector machine algorithm which lead to be difficultly applied extensively,based on prior knowledge of support vector distribution,combined KNN algorithm and 10 fold cross-validation method,a support vector extracted Algorithm was proposed.The proposed algorithm firstly computes k neighboring samples for each original sample,and then computes the proportion q of heterogeneous samples in the neighboring samples.Finally,it selects the samples which meet q greater than threshold value p as support vectors.In this process,the proposed algorithm use 10 fold cross validation method to determine the most appropriate values of k and q.Simulation experiments show that the proposed algorithm can effectively reduce the running time of support vector machine classifiers,while being with a good classification performance.
出处
《华中师范大学学报（自然科学版）》
CAS
北大核心
2014年第3期335-338,共4页
Journal of Central China Normal University：Natural Sciences
基金
河南省2009年高等学校青年骨干教师资助计划项目（2009GGJS-100）.
