SUMMARY
The actual problem of modeling of the multivariate nonstationary time series of economic dynamics is being researched for the purpose of analysis, forecasting and decision-making in financial markets.The proposed approach to the modeling of time series is based on the methodology of multivariate analysis and continuity equation, which relates the probability density function of the state variables of the system with their speeds.Equation of motion of a point in a multidimensional phase space of state variables derived under the assumption that the evolution of the economic system based on the interaction of two factors - the growth and dissipation. It is assumed that the growth rate has a deterministic function, which means that there is a causal link between variables, and the diffusion component of the velocity is proportional to the gradient of the state probabilities in a local point of phase space. In this case the state of the system is determined by multivariate Fokker-Planck equation. On the basis of two-dimensional Fokker-Planck equations is constructed model of the real economic process - trading on the stock exchange. The structure of the model equations of nonlinear responsible paradigm of financial markets and agreed with the results of empirical research. We derive differential equations for the evolution of one-dimensional distributions of prices, trading volume, and spread their moments that are needed to complete the system for the unknown probability density functions.The evolution equations are based on sample data and agreed with the two-dimensional Fokker- Planck equation. Modeling the dynamics and forecasting of trades carried out by numerical integration of the equations of evolution in a sliding window of the sample. The proposed approach to modeling allows the best use of the information contained in the multivariate time series and to obtain high prediction accuracy. Verification of the model performed on the rows indices trading on the Ukrainian stock market.