MVLSC Home · Issue Contents · Forthcoming Papers
A Decision Tree Approach for Predicting Insolvency of SMEs: An Empirical Research in Spain
Andrés Navarro-Galera, Juan Lara-Rubio, Pavel Novoa-Hernández and Carlos Cruz
This paper addresses the problem of insolvency prediction in Spanish SMEs from a binary classification perspective. From a dataset containing more than 325 dependent variables of different natures (including demographic and economic-financial) and 17562 SMEs, a decision tree has been fitted that with a 0.93 accuracy predicts the state of a company and allows it to identify the most relevant variables in such prediction. The results obtained indicate that, in general, insolvency can be predicted from very few financial variables and with an anticipation of more than 10 years before the insolvency event. As a result, it is possible to provide early warnings to SMEs based on the indicators identified.
Keywords: Insolvency, decision tree, SME, binary classification, Pareto optimality, model selection, bankruptcy