Special methodological tools must be used in the analysis of structurally complex economic systems operating with significant indeterminacy, in situations where comprehensive data are lacking or nonfinancial indices must be included. One promising tool is the mathematical apparatus of fuzzy logic. Analytical support for any economic system is based on financial analytic software, usually including tools for the recognition and prediction of financial states with different degrees of indeterminacy. The method proposed in the present article for financial analysis of enterprises consists of two successive procedures: (1) the formulation of membership functions; (2) clustering in the space of financial indicators. Experiments show that the reliability of assignment to one of two clusters (normal state or prebankruptcy state) is comparable for the proposed method and neural network methods. The proposed method retains the benefit of having experts assign membership functions: satisfactory recognition is possible when the data are contradictory and incomplete and the indeterminacy factors are of different kinds.
Bykov et al. (Thu,) studied this question.