Bankruptcy prediction using Logistic Regression and BaggedDecision Trees

By Albert Bezzina, Daniel Farrugia, Ivan Salomone

January 2019

Introduction

Companies having liquidity problems face what is know as financial distress, that is the company’s operating cash flow is insufficient to meet its current liabilities. Financial distress forces companies to take unwanted measures, such as reducing the work force or downsizing operations, in order to alleviate the problem. Failing to emerge from financial distress leads a company to bankruptcy, that is a situation where a company is unable to operate any longer due to its liquidity problems. Bankruptcy affects the social community, business partners, investors, policy makers and the economy as a whole. The high costs of bankruptcy led many researchers to seek means of anticipating it. Such studies date back to the 1930’s when economists used financial indicators to predict business failures. Statistical models for bankruptcy prediction were introduced in the 1960’s. Later in the 1990’s, artificial intelligence and machine learning models took over.

In this assignment we demonstrate the use of Bagged Decision Trees and Logistic Regression models to predict companies going bankrupt based on their financial indicators.

http://archive.ics.uci.edu/ml/datasets/Polish+companies+bankruptcy+data

The End