Neural network data processing for financial insolvency fore-casting of the water supply enterprises

Abstract

Abstract This work is devoted to creation of neural network methods with the increased stability and reliability for bankruptcy forecasting. In this paper, we investigate the performance of four different neural network (NN) methods for bankruptcy prediction of the resource provisioning system enterprises. Two methods based on NN-ensembles of classifiers (boosting-ensembles and bagging-ensembles) and another two tested methods are stand-alone classifiers (multilayer perceptron (MLP) and network of radial basic functions). We defined that bagging-ensembles of neural networks have advantage in comparison with classical - stand-alone classifiers. Developed bagging-ensemble of neural networks is currently used to the sphere of financial monitoring for identification of signs of financial insolvency of the water supply enterprises and assessment of risk of social tension in regions.