摘要
进行基于支持向量机的贷款风险评估研究,在训练向量空间中找到一个分类超平面,使向量分类具有较小的错误率,并获得较强的可扩展能力。从理论分析与实验对比可知,采用遗传算法可使其收敛到全局最优参数,确保支持向量机的分类和扩展的性能达到最优。在平均执行时间小于1 s内获得的高于94.3%正确率的实验结果验证了本算法的正确性与有效性,同时也表明支持向量机在小样本特征空间分类中所具有的优良性能。
This paper describes the use of Support Vector Machine to evaluate the credit risk. It is to find a super-plane in the trained victor to minimize the fault rate and obtain the strong extensibility. With both the theory and experiment, the genetic algorithm demonstrates the converged results which illustrate that the truth rate is higher than 94.3% in the mean execute time less than 1 second, and the Support Vector Machine possess the superior properties in the small sample collections.
出处
《长春工业大学学报》
CAS
2008年第2期196-200,共5页
Journal of Changchun University of Technology
基金
吉林省科技厅基金资助项目(吉科合字第sc0601019)
关键词
贷款风险评估
支持向量机
神经网络
遗传算法
credit risk evaluation
Support Vvector Machine （SVM）
neural network
genetic algorithm
