Interpretable Nonlinear Model for Enterprise Bankruptcy Prediction

• Main Authors: O. Purvinis, R. Virbickaitė, P. Šukys
• Format: Article
• Language: English
• Published: Vilnius University Press 2008-01-01
• Series: Nonlinear Analysis
• Subjects: bankruptcy; prediction; ANFIS; knowledge
• Online Access: http://www.journals.vu.lt/nonlinear-analysis/article/view/14589

The aim of this research was to model bankruptcy dependency of Lithuanian enterprises on their financial ratios and its dynamics over time by the integration of artificial neural networks and fuzzy logic technology using Adaptive Network – based Fuzzy Inference System (ANFIS). We used data from financial reports for three years’ of 230 Lithuanian going and failed enterprises. Input variables used for the ANFIS model training and testing composed of 13 financial ratios of the last year before bankruptcy and 13 variables characterizing changes of that ratios over time. It was checked 1716 subsets of input variables, each subset containing five input variables. This way the ANFIS model and the best subset of predictive variables with minimal training errors was found. Test of that model showed that percentage of right failure and success predictions reached 80 %. Fuzzy rules of the ANFIS were used to construct interpretable rules base, which can be useful for enterprise managers as knowledge for the linguistic diagnosis of failure or financial problems.