Predicting the Relationship between Corporate Financial Information and Credit Rating Using Deep Learning

One of the best ways by which accounting information users reduce investment risks through corporate information is to predict corporate credit ratings, exactly. Thus, this study would utilize deep learning based on an artificial neural network to predict credit ratings relatively exactly based on the corporate financial information. In other words, this study designed a model for the prediction of credit ratings as a neural network consisting of an input layer, two hidden layers and an output layer and evaluated the model through 10-layer cross-validation. Since most preceding studies analyzing the correlation between corporate financial information and credit ratings presupposed the linearity between explanatory variables and dependent variables, there was a limitation for the model’s reflection of the complex real world. To overcome this limitation, this study presupposed a nonlinear activation function based on an artificial neural network and utilized deep learning based on a deep neural network with the increased number of hidden layers. According to the results of this study, the accuracy of the model for the prediction of credit ratings utilizing deep learning was much higher than the random prediction of credit ratings, and through this, it was proven that utilizing deep learning would be useful in predicting credit ratings. Reducing accounting information users’ investment risks through the accurate prediction of credit ratings allows efficient allocation of resources. In addition, this study would provide useful resources for the supervisory institution that supervises the capital market. In other words, when a supervisory institution prepares a system related to corporate credit ratings by providing an accurate model on corporate financial information affecting corporate credit ratings, the results of this study can be utilized as reference data.