欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
基于多元图表示原理,提出散点图可视化分类器.该分类器的基本思想是将数据矩阵映射为散点图,利用散点图表示的紧致性,通过像素图将其转换为图像并利用图像扩展形成分类子空间,最终将多个分类子空间按照组合分类规则构成组合分类器.该分类器集成了图表示技术与图像处理技术,使整个分类过程可视,可实现交互式分类.利用Iris和W ine数据集的实验表明,散点图分类器对训练样本数量敏感度较低,分类性能接近甚至优于目前的主流分类器.
英文摘要：
A novel visual combining classifier based on graphical representation named scatter classifier has been proposed.The basic principle of that is as follows： at first,mapping data matrix to graph by scatter plot,and then using pixel graph convert the graphs space into images space,the sub classifiers are obtained from the images by pixel expanding.At last,different sub classifiers are combined as a combining classifier by combining rule.The classifying process of scatter classifier is visual and interactive,which is the result that the classifier takes advantages of visualization of graphical representation and introduces the image processing technology.A serious of experiments based on Iris dataset and Wine dataset has been done and the experiments show that scatter classifier is insensitive for the size of training set,and the performance of that is close to or even outperformed many popular classifiers.
同期刊论文项目
同项目期刊论文
位置：