머신러닝과 인공신경망을 활용한 수출제조기업 신용등급 예측연구

초록

The purpose of this study is to develop a forecasting model for the credit rating for export manufacturing companies using machine learning and artificial neural network (ANN) after extracting importance variables and compare the performance of each model. The independent variables consists of 16 financial ratios. it is divided into variables selected based on trees model using python scikit-learn module and variables selected by statistical method. An objective variable is credit rating. In order to forecast credit rating, Logistic Regression, XGBoost, Random Forest and ANN are conducted. The result shows that the performance of ANN model is superior to all other models. variables selected based on trees model using python scikit-learn module have higher predictive power than variables selected by statistical method. Debt ratio, equity capital ratio, and operating profit ratio were found to be important variables in credit rating prediction. The performance of ANN is higher in forecasting the credit ratio for export manufacturing companies compared to other models. It was confirmed that the variable selection method based on tree model using the Scikit-learn module could also be a good alternative in a study to minimize the number of independent variables.

키워드