基于多元统计图表示的脸谱图方法和最小均值分类器的思想,提出了最近脸谱图分类器新方法。针对脸谱图不能有效表达数据样本维数的特点,提出了着装脸谱图,并给出了新的图形特征:面积特征和重心特征。针对脸谱图分类器的主观性,提出了利用最近均值分类器实现脸谱图图形特征的自动分类。8种真实数据集的实验结果证明了该方法的有效性。
The nearest Chemoff faces classifier is proposed based on the idea of the Chemoff faces and the nearest mean classifier. The clothes Chernoff faces are proposed to improve the shortcoming of Chernoff faces on non-representing the data dimension. And the two graphical features are proposed: area feature and barycentre feature. The clothes Chernoff faces auto classifier is proposed using the nearest mean classifier and using the proposed graphical features to improve the subjective classified shortcoming of Cher- noff faces. The experiment results of the 8 real data sets prove validity of the new method.