Neuro-Fuzzy Modeling Techniques in Economics

16

Authors: O. Olkhovska

It’s proposed in the article the system of indices of activity of insur-ance companies, which is based on the principles of transparency, comprehensiveness, accessibility, availability of statistical databases, and allows to diagnose insurers, which are stable, and whose finan-cial condition is close to critical. It’s grounded the necessity of further improvement of methodological and instrumental bases of modeling the financial condition of insurance companies, which consist in developing of response system that best meets the requirements of a particular situation. It is the reason of choice of fuzzy logic tools as mathematical basis for the construction of appropriate models. Built economic and mathematical model of bankruptcy diagnosis allows to classify the insurance companies into stable and potential bankrupts taking into account expert knowledge on insurance business, and provides the possibility to optimize their own parameters on real data. It’s proposed the algorithm for tuning the models, based on error-back-propagation method adapted for fuzzy logic models. As a result of tuning on statistical data it’s possible to optimize the parameters of the model, which allow functionally link the input variables (indica-tors of the insurance company) with the value of the result variable (the classes of financially stable insurers or potential bankrupts).

26

Authors: T. Zagorodnya, G. Gavrilyuk

The article describes the features of the account of uncertainty in the judgments of experts through the use of fuzzy preference relations to evaluate the possibility of adjusting the types of work over the product in the structure of the methods of labor. There is shown the features, decomposition, the algorithm of usage of method of paired comparisons to evaluate the possibility of adjusting the types of work over the product. In evaluating the types of work to be correcting or eliminating on the basis of constructed selection criteria, algorithm performs the steps to find a vector selection criteria, execute of ranking and identify the types of work that require correction. As a result it’s suggested an algorithm of estimating the possibility of adjusting the types of work over the product, which is based on the theory of fuzzy sets and gives the possibility to reduce labor costs for execution of methods of labor.

26

Authors: A. Sigal

This work is devoted to the questions of a joint application in economics game theory and fuzzy mathematics. In the paper the process of constructing the fuzzy sets of the most reliable objects (projects) is proposed. To estimate the values of membership functions it’s used the scenario approach and the solution of the corresponding classical or neoclassical antagonistic game that simulates the situation of deci-sion-making under uncertainty, conflict and economic risk. The article describes the methods of solution of neoclassical antagonistic games, based on the classification of information situation. Solutions of specific problems are found.

48

Authors: A. Matviychuk

It is developed in the article the conceptual approach to modeling of financial stability of enterprises which permit to assess the state of the company through diagnosis of its bankruptcy possibility by finding similarities between the performance indicators of this company and enterprises have gone bankrupt and financially stable companies. The methods of theories of fuzzy logic, neural networks and discriminant analysis have been applied for the construction of economic and mathematical models of bankruptcy prediction. The analysis of car- ried out experiments revealed essential misfit of previously known discriminant models to conditions of transformational economy, and also shown relatively high correctness in predicting of enterprises’ bankruptcy with usage of developed by the author economic and mathematical models. All models are built on the same sets of explanatory variables and optimized on the same statistical data of Ukrainian enterprises. The testing of models also was carried out on one statistical sample, which allowed make a comparative analysis and obtain appropriate conclusions about the effectiveness of various mathematical tools to solve the problem of classification of objects of research.