This project explored and compared the performance of new and legacy machine learning models with a history of use in the financial industry for bankruptcy detection.
Bankruptcy is a legal process through which people or corporate entities who cannot repay debts to creditors may seek relief from some or all of their debts. Bankruptcy is often initiated by the debtor, making the use of predictive modeling a major tool in the finance industry to save costs associated with bankruptcy. Early detections allow creditors to analyze their own risk and help mitigate unwanted transactions beforehand.
The detection of bankruptcy benefits creditors, investors, shareholders, partners, and even buyers and suppliers as a measure of financial health, as well as impacting many facets of the financial market, making bankruptcy prediction a longstanding issue within finance, accounting, and management science. The coronavirus pandemic has placed increased stress on financial markets and businesses alike due to many different economic factors, furthering the importance of development of machine learning models in predicting financial indicators such as bankruptcy.
The data was collected from the Taiwan Economic Journal from the years 1999 to 2009. Company bankruptcy was defined based on the definition given by business regulations of the Taiwan Stock Exchange.