We aim to provide a predictive model, specifically designed for the Italian economy, which classifies solvent and insolvent firms one year in advance, using AIDA Bureau van Dijk dataset from 2007 to 2015. We apply a full battery of bankruptcy forecasting models, including both traditional and more sophisticated machine learning techniques, and add to the financial ratios used in the literature a set of industrial/regional variables. We find that XGBoost is the best performer and that industrial/regional variables are important. Moreover, belonging to a district, having a high mark up and a greater market share diminish bankruptcy probability.
Bragoli, D., Ferretti, C., Ganugi, P., Marseguerra, G., Mezzogori, D., Zammori, F., Machine Learning models for bankruptcy prediction in Italy: do industrial variables count?, Working Paper N. 19/3 DIPARTIMENTO DI MATEMATICA PER LE SCIENZE, ECONOMICHE, FINANZIARIE ED ATTUARIALI, Vita e Pensiero, MILANO -- ITA 2019: 3-41 [http://hdl.handle.net/10807/143387]
Machine Learning models for bankruptcy prediction in Italy: do industrial variables count?
Bragoli, Daniela
;Ferretti, Camilla;Ganugi, Piero;Marseguerra, Giovanni;
2019
Abstract
We aim to provide a predictive model, specifically designed for the Italian economy, which classifies solvent and insolvent firms one year in advance, using AIDA Bureau van Dijk dataset from 2007 to 2015. We apply a full battery of bankruptcy forecasting models, including both traditional and more sophisticated machine learning techniques, and add to the financial ratios used in the literature a set of industrial/regional variables. We find that XGBoost is the best performer and that industrial/regional variables are important. Moreover, belonging to a district, having a high mark up and a greater market share diminish bankruptcy probability.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
