第6期 曹苏群，等：最佳鉴别矢量集在无监督模式下的扩展 519 表7采用l-NN和LibSVM分别对经PCA,OSDV和UOSDV降维的Pima_indians._diabetes数据集进行留一法交叉验证所得分 类准确率 Table 7 Leave-one-out rate with 1-NN and LibSVM for reduced-dimension Pima indians diabetes dataset extracted by PCA,OSDV and UOSDV PCA OSDV UOSDV Number of Leave-one-out Leave-one-out Leave-one-out Leave-one-out Leave-one-out Leave-one-out features rate with rate with rate with rate with rate with rate with 1-NN LibSVM 1-NN LibSVM 1-NN LibSVM 1 0.637 0.656 0.682 0.777 0.583 0.660 2 0.664 0.667 0.667 0.772 0.578 0.643 3 0.639 0.618 0.689 0.707 0.565 0.620 0.673 0.637 0.695 0.680 0.615 0.600 5 0.686 0.655 0.685 0.638 0.672 0.645 6 0.682 0.651 0.664 0.642 0.643 0.658 0.680 0.654 0.678 0.655 0.685 0.658 表8采用l-NN和LibSVM分别对经PCA,OSDV和UOSDV降维的Wine数据集进行留一法交叉验证所得分类准确率 Table 8 Leave-one-out rate with 1-NN and LibSVM for reduced-dimension Wine dataset extracted by PCA,OSDV and UOSDV PCA OSDV UOSDV Number of Leave-one-out Leave-one-out Leave-one-out Leave-one-out Leave-one-out Leave-one-out features rate with rate with rate with rate with rate with rate with 1-NN LibSVM 1-NN LibSVM 1-NN LibSVM 1 0.708 0.640 0.944 0.938 0.764 0.702 2 0.719 0.472 0.944 0.961 0.691 0.730 4 0.758 0.461 0.972 0.983 0.736 0.742 6 0.770 0.449 0.983 0.983 0.663 0.657 8 0.770 0.444 0.972 0.978 0.736 0.831 10 0.770 0.444 0.944 0.966 0.888 0.899 12 0.770 0.444 0.803 0.584 0.770 0.685 表9采用1-NN和LibSVM分别对经PCA,OSDV和UOSDV降维的Wdc数据集进行留一法交叉验证所得分类准确率 Table 9 Leave-one-out rate with 1-NN and LibSVM for reduced-dimension Wdbc dataset extracted by PCA,OSDV and UOSDV PCA OSDV UOSDV Number of Leave-one-out Leave-one-out Leave-one-out Leave-one-out Leave-one-out Leave-one-out features rate with rate with rate with rate with rate with rate with 1-NN LibSVM 1-NN LibSVM 1-NN LibSVM 1 0.859 0.852 0.963 0.956 0.931 0.924 2 0.914 0.627 0.968 0.963 0.933 0.935 4 0.916 0.627 0.961 0.961 0.935 0.942 6 0.914 0.627 0.965 0.970 0.937 0.947