Santoni, Valentina (2014) La Previsione dell’insolvenza aziendale: confronto della performance dei modelli Zscore, Logit e Random Forest su un campione di aziende manifatturiere italiane. Doctoral Thesis.
|Full text disponibile come PDF Richiede visualizzatore di PDF come GSview, Xpdf o Adobe Acrobat Reader|
With this study we intend to compare three different methodologies applied for bankruptcy prediction, in order to define which one is the most reliable: Zscore analysis, Logit model and Random Forest.
The aim is to establish if Altman’s Z-score, a widely used tool to evaluate the financial health of a company, is still an efficient methodology to predict bankruptcy or financial stress conditions. Several other forecasting methods have been developed over the years, most of them based on logistic regression.
Here we present a methodology based on a machine learning algorithm (Random Forest) to analyze and predict the bankruptcy of 3.000 Italian manufacturing companies.
We performed the same analysis with Altman's Z-score and Logit model. According to our results, Random Forest obtained the best performance, with a prediction accuracy of 99,85%.
Our results show that applications of machine learning based methods to predict bankruptcy might overcome pre-existing methodologies and be more efficient to identify companies that may become insolvent and unable to repay loans.
I documenti depositati in UnissResearch sono protetti dalle leggi che regolano il diritto d'autore
Repository Staff Only: item control page