摘要
SVM具有优良的学习能力和泛化能力,在应用SVM解决实际问题时,必须对其进行寻优,找到最合适的SVM模型。基于支持向量机模型算法,针对特征选择和核函数及参数选择单独优化的缺陷,提出了将特征选择和核函数及参数选择联合进行的优化方法。仿真实验结果表明,采用联合优化后的模型进行脑中风的预测诊断,准确率达94.3%。与单独优化相比,不仅提高了分类性能,而且缩短了模型的判别时间。
SVM has good generalization ability and learning ability. In the practical applications, it must be optimized to find the most appropriate model. Based on support vector machine model algorithm, a method was put forward to optimize the model with the joint o f feature selection and kernel function and parameters. It solved the defects of separate optimization feature and kernel function and parameters. Simulation results show that the joint optimized models predict stroke diagnosis, its accurate rate is 94.3%. Comparing with a separate optimization methods, it not only improves the classification performance, but shortens the discriminant time.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第12期3833-3836,共4页
Journal of System Simulation
基金
山西省“十一五”规划课题(GH-06211)
山西医科大学创新基金(01200824)
关键词
支持向量机
核函数
模式识别
脑中风
support vector machines
kernel
pattern recognition
Apoplexy