| Session | ||
Session 3.24: Forecasting of default risk: machine learning application on SMEs financial data
| ||
| Presentations | ||
Forecasting of default risk: machine learning application on SMEs financial data Department of Economics, University of Molise, Italy Despite numerous contributions on the topic, the study of the dynamics that influences the risk of SME insolvency still finds remarkable interest. In recent years, the use of machine learning algorithms in this field has increased the predictive accuracy of credit risk models. Using a set of fourteen indicators derived from a proprietary dataset, our study compare the predictive effectiveness of different machine learning models, currently widely used in the literature and in credit risk applications, through the calculation of specific evaluation indicators (e.g., accuracy, precision, F1 score, ROC curve (AUC), precision/recall curve). In addition, we provide useful information on the role of financial and accounting indicators in providing warning lights to entrepreneurs and managers to anticipate and manage potential default risks through the implementation of a Feature Importance Analysis. | ||