A Bankruptcy Prediction Model Using Random Forest

Abstract

Bankruptcy Prediction is very important for an organization as well as decision makers such as financiers and investors. Selection of dataset for training the prediction model, the Machine Learning tool used for prediction and various other factors are essential in building an efficient prediction model. The dataset includes financial ratios as attributes that are derived from the financial statements of various companies. The most influencing ratios that are required for predicting bankruptcy are selected on the basis of the Genetic Algorithm which filters out the most important ones from different existing bankruptcy models. These ratios of different companies are fed as an input to train the model being implemented in R. The prediction algorithm used is Random Forest, which will enable us to differentiate between bankrupt and non-bankrupt companies.