Abstract

The prediction of corporate financial failure, crucial for the prevention and mitigation of economic downturns in a national economy, requires the categorization of healthy and unhealthy companies. This study examines the case of Serbia and applies multivariant statistical methods and specific artificial neural network architectures—the self-organizing map —to assess the corporate financial health of various companies. Financial ratios drawn from corporate balance sheets become the independent variables in a multivariate discriminant analysis . These financial ratios and the discriminant Z-score in the MDA form the input for the SOM, which creates a hybrid MDA-SOM model that is capable of predicting corporate financial insolvency. The experimental results of this research correctly estimate company financial health in 95% of cases. These are reliable predictions that are comparable with similar studies in other countries