기업구조조정 혁신을 위한 선제적 한계기업 예측모형에 관한 연구

Since the introduction of the work-out system, which can be called the basis for corporate restructuring, the management system for corporate insolvency are shifting. But it has the limitation that they are biased toward follow-up management. Therefore, this study aims to build a preemptive prediction model for marginal companies through preliminary analysis of deteriorating factors in business. As of 2015-2018, we have collected 477,458 data on about 150,000 companies per year and selected various financial ratios and macroeconomic variables as input variables. In addition, we intended to propose a robust model for the prediction of companies that are showing the signs of insolvency by comparing performance with methodologies using various machine learning and traditional statistical models. As a result of the experiment, k-NN(k-Nearest Neighbor) algorithm has established a robust model for both normal and marginal companies. In particular, for marginal companies, the prediction performance of k-NN algorithm is not significantly different from year to year unlike other methodologies, thus it is the most suitable for the purpose of this study to build a preemptive prediction model.