Applying machine learning techniques to enhance the predictive power of the Altman Z-score model in European Union companies

: an empirical study

Abstract

Predictive bankruptcy refers to using statistical models and financial analysis techniques to determine the likelihood of a company or organization going bankrupt. This can help investors, lenders, and other stakeholders make informed decisions about their exposure to financial risk. Altman's Z-Score model is considered an effective tool for predicting bankruptcy and is widely used. However, the model relies only on a few financial ratios as covariates and does not include other variables that may be relevant to predicting bankruptcy. This thesis starts by analyzing the predictive capabilities of the Altman Z-Score model in a set of European Union countries for private non-manufacturing companies. This is followed by testing if modern machine learning technics can achieve a better prediction when predicting bankruptcy. In the end, which features between a company's financial statement and the ones that Altman used in his model are the most important to consider while predicting bankruptcy. Altman Z-Score model showed poor prediction capability for bankrupt companies. In contrast, the machine learning models showed increased predictive capabilities by conjugating variables from the Altman Z-Score model and company financial statements. It is possible to build a model with higher precision.

Date of Award	3 May 2023
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Nicolò Bertani (Supervisor)