Data Mining for Predicting and Finding Factors of Bankruptcy

ABSTRACT

The financial crisis, since the pandemic outbreak due to COVID-19, the dissemination and invasive systemic risk in the global financial environment have drawn the attention to organizations' solvency monitoring methods. Inevitably, in this paper we have looked at the both bankruptcy prediction and the factors that lead to bankruptcy. The dataset for this study was acquired from Kaggle, which was based on the Taiwan Economic Journal, from 1999 to 2009. The corporate statutes of the Taiwan Stock Exchange were utilized to determine a company's bankruptcy. It was a highly imbalanced dataset having 220 Non-bankrupt and 6,599 bankrupt data. We have used Random Forest, Support Vector Machine, Artificial Neural Network, XGBoost, and LightGBM classifiers regarding bankruptcy prediction. On the other hand, to find the factors that lead to bankruptcy, we did an empiric analysis for which we calculated fourteen statistical values of both bankrupted and non-bankrupted features and saw their cosine similarities. These factors will help any financial company to plan its financial ratios for preventing bankruptcy. Here we got the best performance from the Artificial Neural Network with 98.64% accuracy. And we found four factors that were responsible for the bankruptcy in our dataset. Here, the factors determining bankruptcy are crucial because by finding these factors and the likelihood of bankruptcy, companies can take the necessary steps to plan their financial ratios and ensure the solvency of their businesses. © 2022 IEEE.