摘要：This paper analyses the possibilities of prediction ofthe enterprise bankruptcy, applying neural network. Theprediction results of the failed enterprises are comparedto the prediction results of the profitable enterprises. Inthis way we can see the peculiarities and reliability ofneural network usage for the bankruptcy diagnosis.Lithuanian enterprises work in different conditionsthan other foreign enterprises, but all introduced modelswere based only on foreign enterprises, so their applicabilityfor the diagnostics of bankruptcy remainsdisputable. Estimations of enterprises for the bankruptcyin definite time, calculated from “Z” scores ofdifferent authors, are different so they have to be analysedtaking into account of their changing tendencies.This article discuses the application of neural networksto analyse the possibility of enterprise bankruptcy.The classification of neural networks, estimationof the number of hidden layers and their size, themethods of training are described in special scientificliterature.Perceptrone neural network was constructed of 3layers. To train it the backpropagation method wasused. The algorithms of training and the programmes toimplement them require a lot of samples of enterprises –over ten times more than inputs of the enterprise state.To train the network following indicators were used: theindicator of net profitability of assets, coefficient ofshort-term solvency; debt ratio; ratio of short-term liquidityof the years 1998-2001. The authors had data of13 enterprises, so they increased the number by includingthe same enterprises in the list several times. In thisway 284 enterprises were obtained: 161 failed and 123profitable.By training various networks with different inputs itwas researched what indicators of the enterprise werethe best to forecast the bankruptcy. Therefore the neuralnetwork was trained in the optimisation mode. The programmeused different combinations of inputs andchecked 408 different versions of the neural networks.As a result, all the used inputs could forecast bankruptcy,except the profitability of assets of the year 2000and the short-term liquidity ratio of the year 1998.According to the small amount of enterprises (8profitable and 5 insolvent) and their 4 financial ratiosused to train the neural network, the percentage of theright diagnosis is 84. But when increased the number ofenterprises to 284 (written the same enterprises a fewtimes in the same list), the results of right diagnosisrose to 92 per cents. It is good result of the method of the neural network prognosis. The research is beingcontinued.
其他摘要：This paper analyses the possibilities of prediction ofthe enterprise bankruptcy, applying neural network. Theprediction results of the failed enterprises are comparedto the prediction results of the profitable enterprises. Inthis way we can see the peculiarities and reliability ofneural network usage for the bankruptcy diagnosis.Lithuanian enterprises work in different conditionsthan other foreign enterprises, but all introduced modelswere based only on foreign enterprises, so their applicabilityfor the diagnostics of bankruptcy remainsdisputable. Estimations of enterprises for the bankruptcyin definite time, calculated from “Z” scores ofdifferent authors, are different so they have to be analysedtaking into account of their changing tendencies.This article discuses the application of neural networksto analyse the possibility of enterprise bankruptcy.The classification of neural networks, estimationof the number of hidden layers and their size, themethods of training are described in special scientificliterature.Perceptrone neural network was constructed of 3layers. To train it the backpropagation method wasused. The algorithms of training and the programmes toimplement them require a lot of samples of enterprises –over ten times more than inputs of the enterprise state.To train the network following indicators were used: theindicator of net profitability of assets, coefficient ofshort-term solvency; debt ratio; ratio of short-term liquidityof the years 1998-2001. The authors had data of13 enterprises, so they increased the number by includingthe same enterprises in the list several times. In thisway 284 enterprises were obtained: 161 failed and 123profitable.By training various networks with different inputs itwas researched what indicators of the enterprise werethe best to forecast the bankruptcy. Therefore the neuralnetwork was trained in the optimisation mode. The programmeused different combinations of inputs andchecked 408 different versions of the neural networks.As a result, all the used inputs could forecast bankruptcy,except the profitability of assets of the year 2000and the short-term liquidity ratio of the year 1998.According to the small amount of enterprises (8profitable and 5 insolvent) and their 4 financial ratiosused to train the neural network, the percentage of theright diagnosis is 84. But when increased the number ofenterprises to 284 (written the same enterprises a fewtimes in the same list), the results of right diagnosisrose to 92 per cents. It is good result of the method of the neural network prognosis. The research is beingcontinued.
关键词：bankruptcy diagnosis; financial indicators of enterprise; neural networks.
其他关键词：bankruptcy diagnosis; financial indicators of enterprise; neural networks.
