摘 要:提出了一种用支持向量机(SVM)权重向量解决高维对象分类的方法,并结合云理论建立了基于SVM权重向量的云分类器。采用云模型建立训练集的各属性模型,分类模型由属性模型集成得到,属性权重根据SVM权重向量得到,属性权重越大,其对分类的贡献越大;反之,越小。将新分类器与云模型分类器对积雨云、卷云和卷层云进行分类模拟实验,新分类器的分类准确度比后者总体提升了, 经过交叉验证, 结果表明新分类器性能稳定。
New cloud classifier based on SVM weight vector
ZHU Jie1,QIN Liang-xi1,LONG Wei-zhe1,SU Yong-xiu2
(1.School of Computer, Electronics & Information, Guangxi University, Nanning 530004, China;2.Guangxi Institute of Meteorological Disaster Mitigation, Nanning 530022, China)
Abstract:This paper presented a method of support vector machine weight vector to solve the problem of high-dimensional objects classification, and built cloud classifier based on the cloud theory and the SVM weight vector.Set up the attribute model of training by cloud model.Integrated classification model by every attribute model, and attributed weight came from the SVM weight vector. The larger the weight of an attribute was, the more it would make the contribution to classification. On the contrary, it would reduce the effect of classification. The new classification algorithm and the cloud classifier were applied to classify the radiance profiles as cumulonimbus, cirrus clouds, or cirrostratus clouds. The experimental results show that it must can improve the classification accuracy of spatial data in the overall performance than the latter.This cross validation to prove that the performance of the new classifier is pretty steady. ......