摘要
通过支持向量机(SVM)对客车车型的长,宽,高,宽长比等7个特征进行特征选择,得到的准确率最高的子集是长、宽、高、宽长比、宽高比,以它作为样本特征进行分类.对客车的4类车型进行分类,每类车型选择80个样本,50个样本进行训练,30个样本进行预测,结果表明:对1类车型的分类准确率可达到100%,对2类和4类车型可达到96%以上,对3类车可达到93%以上.得到了比选用长、宽、高作为特征进行分类更优的结果.然后运用加入参数寻优的SVM对客车的4类车型进行分类,并加以比较.基于高斯函数的特性,两次用到SVM进行机器学习时,核函数均选用RBF核函数.
This paper use the SVM to choose the features which belong to Passenger vehicles, the feature include length,width,height and the ratio of them and so on.Then we get the best accuracy of the subset is consisted of length,width,height,width divided by length and width divided by height.Let the subset＇s features as the input and classify the samples.In this paper,we classify the passenger vehicles type which have 4 class,each type have 80 samples ,we use 50 samples for training and 30 samples for predicting.The results show that the first type classification accuracy can reach 100%,the second and third type can exceed 96% and the fourth type can exceed 93%,the result is better than the result which choose length,width,height as the features,then use the SVM which join the parameters optimization to classify the passenger car include 4 type and compare the result.Based on the characteristics of the gaussian function ,we have twice to use the SVM for machine learning,and choosing RBF Kernel Function as the kernel function.
出处
《数学的实践与认识》
CSCD
北大核心
2012年第18期190-194,共5页
Mathematics in Practice and Theory
基金
2009年山西省自然科学研究基金(2009011018-3)
关键词
特征选择
C—SVC
核函数
参数寻优
feature selection
C-SVC
kernel function
parameters optimization
