微型机与应用
微型機與應用
미형궤여응용
MICROCOMPUTER & ITS APPLICATIONS
2014年
9期
73-76
,共4页
信用评分%分类回归树%Bayes判别%神经网络
信用評分%分類迴歸樹%Bayes判彆%神經網絡
신용평분%분류회귀수%Bayes판별%신경망락
credit scoring%CART%bayes decision%neural network
在分类回归树模型、Bayes 判别模型和神经网络模型3种评分模型的基础上,构建了相应的信用评分准则,提供更精准的信用卡信用信息评估参数,以便能更好地判别是否接受信用卡申请人的申请。实验结果表明,扩展神经网络评估模型具有相对较高的预测验证准确性,即CART=68.11%, Bayes=67.83%, NN=69.27%。
在分類迴歸樹模型、Bayes 判彆模型和神經網絡模型3種評分模型的基礎上,構建瞭相應的信用評分準則,提供更精準的信用卡信用信息評估參數,以便能更好地判彆是否接受信用卡申請人的申請。實驗結果錶明,擴展神經網絡評估模型具有相對較高的預測驗證準確性,即CART=68.11%, Bayes=67.83%, NN=69.27%。
재분류회귀수모형、Bayes 판별모형화신경망락모형3충평분모형적기출상,구건료상응적신용평분준칙,제공경정준적신용잡신용신식평고삼수,이편능경호지판별시부접수신용잡신청인적신청。실험결과표명,확전신경망락평고모형구유상대교고적예측험증준학성,즉CART=68.11%, Bayes=67.83%, NN=69.27%。
On the base of three scoring models, which are classification and regression tree model, Bayes model and neural network model, this paper constructs the relevant credit scoring criterion and provides more accurate parameters of credit information in order to get better judge whether to accept credit card applications. Results show that modified neural network model has higher accuracy, the CART is 68.11%, Bayes is 67.83%, NN is 69.27%.