عنوان مقاله [English]
Objective: Convolutional neural networks (CNN)are being applied to identiﬁcation problems in a variety of ﬁelds, and in some areas are showing higher discrimination accuracies than conventional methods. Applications of convolutional neural networks to ﬁnancial analyses have only been reported in a small number of studies because it seems to be that convolutional neural networks are more suitable for application to images and less suitable for general numerical data including ﬁnancial state ments. Hence, in this research, an attempt is made to apply a convolutional neural network to the prediction of corporate bankruptcy and comparise to traditional models.
Methods: The ﬁnancial statements ratios has been choiced 66 companies that have been delisted from the Iran Stock Market due to de facto bankruptcy as well as the ﬁnancial statements of 66 listed companies over 2000 to 2019 ﬁnancial periods. In this method, a set of ﬁnancial ratios are derived from the ﬁnancial statements and represented as a grayscale image. The image generated by this process is utilized for training and testing a convolutional neural network. The images for the bankrupt and continuing enterprises classes are used for training the convolutional neural network based on GoogLeNet. In this research, it generates as many ﬁnancial ratios as possible from the ﬁnancial statements of each company in each ﬁscal year and express the set of ratios as a single grayscale image. To achieve this, each ﬁnancial ratio is made to correspond to a speciﬁc pixel position (x,y-coordinates) and the brightness value of that pixel is set based on the value of the corresponding ﬁnancial ratio. The images generated with this process are then used as input to train the CNN.
Results: The findings shows, in prediction of going concern of firms, Convolutional neural network has predicted with 50 percent of precision. This means that 50 percent of continues firms and 50 percent of bankrupt firms has been predicted precisely. In other hands, to strengthening the findings, three traditional model comprise to Convolutional neural networks and findings show that precise in traditional models is higher to Convolutional neural networks.
Conclusion: Convolution networks can predict bankruptcy by inputting ﬁnancial ratios as an image. Convolution-network-based bankruptcy prediction does not outperforms traditional models as altman (1983),Springate (1978) and Zmijewski (1984). However, unlike some conventional methods, it is hard to know from the proposed method which of the ﬁnancial ratios has a stronger impact on bankruptcy prediction. Therefore, we have to admit that the proposed method is not suitable for the purpose of investigating the causes of bankruptcy.