Support Vector Machine( SVM)involves the storage and calculation of the matrix. Therefore,the huge space and time involved limit the application of SVM. In order to improve the training speed and save storage space,this paper presented a method to reduce the number of the training data. Firstly,the method re-moved most of the non-boundary samples and keeps only a few samples as training data by two steps reduc-tion. Then,KNN was used to remove some noise. Breast-Cancers of UCI standard data sets were employed to evaluate our method. Twenty-five data were removed from the set of support vectors and the training time was decreased by 7 ms,while the accuracy of classification has improved. The experimental result shows that the proposed method can effectively reduce the number of training data,and decrease the training time with the increase in the accuracy of classification.%支持向量机算法求解会涉及矩阵的存储与运算，因此算法的时空复杂度较大，这些不足之处限制了支持向量机的应用。为提高支持向量机的训练速度，缩短训练时间，提出一种样本约简方法。该方法通过两次样本约简，剔除掉大部分非边界样本，保留少数且有效的样本作为训练集。然后，采取KNN算法去除约简后训练集中的孤立点和噪音点。最后，对 UCI 标准数据集中的Breast-Cancers数据进行实验，支持向量减少了25个，训练时间减少了7 ms，而准确率却得到了提高。实验结果表明，在保证预测精确度的前提下，该算法能够有效进行样本约简，缩短训练时间。
展开▼
