CAJ | 학술논문

有向无环图支持向量机( DAG-SVM)对于N类别分类问题,会构造N ×( N-1)/2个支持向量机分类器(为每2个类构造一个支持向量机),DAG-SVM可能出现由于节点选择不佳而导致整个分类器分类结果较差的情况。为此,提出一种改进的DAG-SVM。通过为每一层建立备选节点集合进行节点选择,选取下层备选节点集合中训练分类精度最高的一个节点组合作为当前层节点的下层节点,从而优化DAG-SVM的拓扑结构。实验结果表明,与已有的DAG-SVM,1-vs-1 SVM,1-vs-a SVM方法相比,该方法的分类精度较高。
유향무배도지지향량궤( DAG-SVM)대우N유별분류문제,회구조N ×( N-1)/2개지지향량궤분류기(위매2개류구조일개지지향량궤),DAG-SVM가능출현유우절점선택불가이도치정개분류기분류결과교차적정황。위차,제출일충개진적DAG-SVM。통과위매일층건립비선절점집합진행절점선택,선취하층비선절점집합중훈련분류정도최고적일개절점조합작위당전층절점적하층절점,종이우화DAG-SVM적탁복결구。실험결과표명,여이유적DAG-SVM,1-vs-1 SVM,1-vs-a SVM방법상비,해방법적분류정도교고。
Directed Acyclic Graph Support Vector Machine ( DAG-SVM ) is a novel algorithm of multi-class classification. For an N-class classification problem, DAG-SVM can construct N × ( N -1 )/2 SVM classifiers ( one classifier for a pair of classes ) but DAG-SVM may behave poor due to the poor selection of nodes, concerning the situation raised before,the new method is proposed and the nodes selection is to establish alternative sets of nodes for every layer,and it chooses the nodes group which gets the highest training classification accuracy as the lower layer of current layer form the alternative sets of nodes, so as to optimize the topology structure of DAG-SVM. Experimental results show that compared with other methods like DAG-SVM,1-vs-1 SVM and 1-vs-a SVM,the classification accuracy of this method is high.