摘要
分析了模糊K-近邻方法FKNN,研究了模糊最小最大神经网络FMMNN分类方法,然后深入分析了支持向量机SVM原理,并在此基础上给出了一种改进的径向基核函数。基于IRIS数据,进行了计算机仿真实验,结果表明,改进的SVM方法分类性能比模糊最小最大神经网络与模糊K-近邻算法的分类性能更好,且运算时间更短,更易于实时实现。
Pattern classification is not only an important aspect in pattern recognition but also the key technique in transaction of other questions. Firstly the method of FKNN is analyzed. Secondly the method of FMMNN is also studied. Thirdly studying the theory of SVM, this paper presents an ameliorated RBF kernel function. Finally, the simulation is processed based on the IRIS data and the results are compared with each other. These results show that the performance of the ameliorated SVM outperforms those of the FMMNN and FKNN, with shorter operation time and more suitable for real-time implementation.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2007年第1期99-102,共4页
Systems Engineering and Electronics
基金
国家自然科学基金(60402032)资助课题
关键词
目标识别
模式分类
支持向量机
模糊K-近邻
神经网络
target recognition
pattern classification
neural network real-time implementation. support vector machine
fuzzy K-nearest neightork
