Applying a probabilistic neural network to hotel bankruptcy prediction

Abstract

Using a probabilistic neural network and a set of financial and non-financial variables, this study seeks to improve the ability of the existing bankruptcy prediction models in the hotel industry. Our aim is to construct a hotel bankruptcy prediction model that provides high accuracy, using information sufficiently distant from the bankruptcy situation, and which is able to determine the sensitivity of the explanatory variables. Based on a sample of Spanish hotels that went bankrupt between 2005 and 2012, empirical results indicate that using information nearer to bankruptcy (one and two years prior), the most relevant variable is EBITDA to Current Liabilities, but using information further from bankruptcy (three years prior), Return on Assets is the best predictor of bankruptcy.