NeuroEvolution of Augmenting Topologies for predicting financial distress: A multicriteria decision analysis

Abstract This paper introduces a novel approach based on NeuroEvolution of Augmenting Topologies to early predict financial distress of Tunisian companies using an important number of inputs. Our sample covers the period of the Jasmin Revolution that led to an increase of the number of bankruptcies, making early previsions even more difficult. Furthermore, we aim to identify the factors that explain financial distress as our approach does not need a process of preselection. All the financial ratios will be used as inputs for the model, and only the ratios with the highest threshold value will remain in the final classifier. To test the accuracy of our model, we will compute a linear discriminant analysis to compare the previsions of the two approaches. Results show that our approach outperforms the traditional one.