ScholarWorks@경상국립대학교

초록

This study aims to develop an optimal prediction model for stock delisting in companies listed on the KOSPI and KOSDAQ markets of the Korea Exchange. To enhance the predictive performance of the models, we collected a dataset incorporating various financial ratios and macroeconomic indicators as additional variables, providing a better reflection of the economic conditions at the time. The dataset consisted of financial ratios and macroeconomic indicators from delisted or managed companies from 2014 to 2021. We constructed stock delisting prediction models using individual and ensemble machine learning algorithms, as well as one deep learning algorithm. Additionally, we adopted processes for adjusting classes and utilizing GridsearchCV to further improve the model’s performance. As a result, we identified significant factors influencing a company’s stock delisting risk and found the optimal prediction model by comparing the performance of machine learning algorithms to the deep learning algorithm. We hope these findings offer valuable insights that can assist investors and regulatory authorities in evaluating companies’ financial stability and identifying potential stock delisting risks.

키워드