摘要
阐述一种用微机实现白细胞自动分类的方法。系统采用二分树分类器实现对6种 白细胞分类。对确定二分树分类器的结构、细胞特征选择和结点分类器等问题作了详 细讨论;提出两种适用于微机系统的特征选择方法——相关分析法和二步选择法;给 出不同情况下系统的分类准确率。结果表明.本文提出的方法在训练集上获得的分类 准确确率达93%,考试集上的分类准确率为86%.
A method of automated classification of white blood cells by microcompu-ter is presented in this paper. The system is based on binary tree classification problems of the tree building, the feature selection and the node classifier are discussed in detail. Two kinds of methods of feature selection suitable for microcomputer systems-the correlation analysis method and the two-step- choosing method- is proposed. The classification accuracy obtained is up to 93 percent in the training sets and 86 percent in the testing sets by using weighted K-NN classifier on each node. The results of using different methods and node classifiers are given for comparison.
出处
《大连理工大学学报》
EI
CAS
CSCD
北大核心
1989年第2期217-222,共6页
Journal of Dalian University of Technology
关键词
白细胞
二分树
分类器
微机
分类
leukocytes
binary trees
feature extraction: feature selection, weight (mathematics) /K-nearest neighbor classifier