CAJ | 학술논문

利用因子分析法建立了我国中小企业信用评价财务指标体系，据此构建了单隐层结构的BP神经网络，并以2006年100家ST和非ST上市公司为训练样本，详细考察了隐含层结构和网络训练函数对BP神经网络辨识性能的影响，最终得到了具有较快收敛速度和较高准确率的BP神经网络。在此基础上，以2006年财务数据为研究样本，对102家上市中小企业进行了信用评价。结果表明，各中小企业之间的信用评价结果存在较大的异质性，说明采用BP神经网络进行信用评价具有较高的可操作性。
이용인자분석법건립료아국중소기업신용평개재무지표체계，거차구건료단은층결구적BP신경망락，병이2006년100가ST화비ST상시공사위훈련양본，상세고찰료은함층결구화망락훈련함수대BP신경망락변식성능적영향，최종득도료구유교쾌수렴속도화교고준학솔적BP신경망락。재차기출상，이2006년재무수거위연구양본，대102가상시중소기업진행료신용평개。결과표명，각중소기업지간적신용평개결과존재교대적이질성，설명채용BP신경망락진행신용평개구유교고적가조작성。
Based on factor analysis, this dissertation established a financial credit evaluation indexes system of small and medium-sized enterprises （SMEs）, and constructed BP neural network with single hidden layer. Then, adopting 50 ST companies and 50 non-ST companies as analyze samples, through investigating the effect of hidden layer structures and training functions on the identification capability of BP neural network, this dissertation got a BP neural network with quick convergence speed and high accuracy. At last, using the financial indexes of 2006, this dissertation evaluated the credit of 102 listed SMEs. The result showed that there are relatively large disparities among the credit of the 102 listed SMEs,which proved that the BP neural network is very suitable for the credit evaluation.