摘要
研究表明支持向量机集成方法可以提高分类精度,但是目前所用的基于最多投票原则的集成策略无法评价单个支持向量机分类器的输出重要性。针对这个问题,本文建立一种基于五级分类的支持向量机集成方法,该方法具有四个因子输入,一个衡量商业银行信用风险的输出,该方法考虑了各子分类器的分类结果和各子分类器判决对最终决策的重要程度,利用L ibsvm对某商业银行信贷的176组样本数据进行实证分析,结果表明本文提出的方法比其他分类方法的分类精度高,证实了该方法的可行性和有效性,为银行建立可靠的评估系统提供了依据。
Support vector machines （SVMs） ensemble has been proposed to improve classification performance recently. However, currently used fusion strategies do not evaluate the importance degree of the output of individual component SVM classifier when combining the component predictions to the final decision. A SVMs ensemble method based on fiveclass is presented in this paper to deal with this problem. This method is 4 factor inputs and one output measuring the credit risk of commercial banks. It aggregates the outputs of separate component SVMs with importance of each compo- nent SVM. A substantiation analysis has been made with a comercial bank＇ s 167 groups of sampled data using Libsvm. The experimental result shows that the proposed method outperforms other methods in classification accuracy. It demonstrates the proposed method is highly accurate and feasible, which is useful for providing a sound credit assessment system for commercial banks.
出处
《预测》
CSSCI
北大核心
2009年第4期57-61,共5页
Forecasting
基金
国家自然科学基金资助项目(70773029)
国家教育部博士点基金资助项目(20050213037)
新世纪优秀人才支持计划资助项目(08-0171)
黑龙江省青年科学基金资助项目(QC04C25)
关键词
信用风险评估
支持向量机集成
预测
credit risk evaluation
support vector machines ensemble
prediction
