Financial Risk Forecasting Using Neuro-Fuzzy Approach

The dissertation is devoted to the decision of the problems directed to the development of methods, models and algorithms for solving forecasting problems of financial risks under conditions of uncertainty, and for a complex of the problems related with it. These are fuzzy mathematics operations, the solving of linear algebraic equations system with fuzzy numbers (variables) in the neural network logic basis. This allows essentially raising the level of support of decision-making in the conditions of uncertainty and, as consequence from this, control efficiency. As a result of this, the mechanism of fuzzy conclusion in neural network logic basis is studied, namely it was suggested to use a connectional neural network, realizing the technique of fuzzy conclusion particularly, and fuzzy modeling in general. The problem of optimal borrower selection is realized in the program shell of the MATLAB/Fuzzy Sets Toolbox on current data.