A comparative analysis of machine learning models for corporate default forecasting

Abstract

This study examines the potential benefits of utilizing machine learning models fordefault forecasting by comparing the discriminatory power of the random forest and XGBoostmodels with traditional statistical models. The results of the evaluation with out-of-timepredictions show that the machine learning models exhibit a higher discriminatory powercompared to the traditional models. The reduction in the sample size of the training datasetleads to a decrease in predictive power of the machine learning models, reducing the differencein performance between the two model types. While modifications in model dimensionalityhave a limited impact on the discriminatory power of the statistical models, the predictive powerof machine learning models increases with the addition of further predictors. When employinga clustering approach, both traditional and machine learning models exhibit an improvement indiscriminatory power in the small, medium, and large firm size clusters compared to theprevious non-clustering specifications. Machine learning models exhibit a significantly higherability to classify micro firms. The findings of this research indicate that the machine learning models exhibit superior discriminatory power compared to the traditional models across thedifferent specifications. Machine learning models can be used to forecast the potential impactof corporate default of non-financial micro cooperations on the Portuguese labour market byestimating the number of jobs at risk.

Date of Award	9 May 2023
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Eva Schliephake (Supervisor)