금융 데이터 및 텍스트 데이터를 활용한 금융 기업 조기 경보 모형 개발 : 부실은행 예측을 중심으로

Abstract

The early warning model, one of the risk management models for financial companies, was introduced to detect inadvance the capital adequacy crisis of financial institutions. The existing early warning models have mainly usedpredictive models with structured data-based variables such as various macroeconomic indicators and enterpriseinternal indicators. The purpose of this study is to further improve the performance of the early warning models byapplying customer complaints data and news articles related to individual financial institutions to machinelearning-based model. As a result of applying this technique to actual data over the period of 2001 to 2017, themethodology proposed in this study has demonstrated a performance improvement up to 20%p compared to theexisting methodology in certain evaluation metrics.