摘要
针对传统分类方法只采用一种分类器而存在的片面性,分类精度不高,以及支持向量机分类超平面附近点易错分的问题,提出了基于支持向量机(SVM)和k-近邻(KNN)的多特征融合方法。在该算法中,设样本集特征可分为L组,先用SVM算法根据训练集中每组特征数据构造分类超平面,共构造L个;其次用SVM-KNN方法对测试集进行测试,得到由L组后验概率构成的决策轮廓矩阵;最后将其进行多特征融合,输出最终的分类结果。用鸢尾属植物数据进行了数值实验,实验结果表明:采用基于SVM-KNN的多特征融合方法比单独使用一种SVM或SVM-KNN方法的平均预测精度分别提高了28.7%和1.9%。
The traditional classification methods only use one single classifier,which may lead to one-sidedness,low accuracy,and that the samples nearby the Support Vector Machine(SVM) hyperplanes are more easily misclassified.To solve these problems,the multi-feature fusion method based on SVM and K-Nearest Neighbor(KNN) classifiers was presented in this paper.Firstly,the features were divided into L groups and the SVM hyperplanes were constructed for each feature of training set.Secondly,the testing set was tested ...
出处
《计算机应用》
CSCD
北大核心
2009年第3期833-835,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(10871022)
关键词
支持向量机
K-近邻
多特征融合
后验概率
Support Vector Machine(SVM)
K-Nearest Neighbor(KNN) algorithm
multi-feature fusion
inverse probability