지도학습을 통한 기업의 파산 예측 요인 분석: NCA를 중심으로

The causes of firm’s bankruptcy are diverse and complex, with financial information often serving as critical indicators for prediction. Extant studies on developing bankruptcy prediction models have utilized financial ratios and employed dynamic approaches and statistical techniques to enhance predictive accuracy. Recently, AI and machine learning methods have been introduced to further advance these models. However, there are some limitations in systematically analyzing the interactions between variables and clearly identifying their importance. In this study, therefore, we explored financial ratios that significantly influence firm’s bankruptcy prediction and prioritized these factors based on their level of impact to identify elements that improve the predictive accuracy of bankruptcy models. For this, we collected data containing bankruptcy information on Taiwanese firms and applied statistical techniques and NCA using R and Python. The findings revealed that indicators such as "total revenue to total expense ratio," "net asset value per share," "profit before tax to paid-in capital ratio," and "net asset growth rate" are variables with substantial influence on bankruptcy.