Modeling the Stock Market Dynamics by Takagi–Sugeno Fuzzy Inference Systems

In this paper we evaluate the parameters and study the applicability of the Takagi – Sugeno models to describe the dynamics of the main indices of the Moscow stock market: MI CEX, RTS indices and the oil and gas industry index. Review of TS-models for forecasting foreign stock indices and stock prices is given. TS-models represent a generalization of classical econometric approaches. It is achieved using systems of fuzzy rules. Any Takagi – Sugeno model can be considered as a modification of some linear econometric model. Results from the approximation theory show that a Takagi – Sugeno model can approximate any nonlinear econometric model. In this paper, we construct TS-models for Russian stock indices. The fuzzy clustering method is used to find the membership functions. The coefficients of the linear equation in each fuzzy rule are found using the Sugeno – Kang procedure based on the least squares method. Calculations show that the Takagi – Sugeno model in all cases gives a decrease in the forecast error in comparison with the unmodified linear model. In some examples, the forecast error decreases roughly in 4 times.