The Influence of Variable Selection Methods on the Accuracy of Bankruptcy Prediction Models

Over the last four decades, bankruptcy prediction has given rise to an extensive body of literature, the aim of which was to assess the conditions under which forecasting models perform effectively. Of all the parameters that may influence model accuracy, one has rarely been discussed: the influence of the variable selection method. The aim of our research is to evaluate the prediction accuracy of models designed with various classification techniques and variables selection methods. As a result, we demonstrate that a search strategy cannot be designed without considering the characteristics of the modeling technique and that the fit between the variable selection method and the technique used to design models is a key factor in performance.