Multi-regression Forecast in Stochastic Chaos

Abstract: This paper addresses the challenge of short-term forecasting for processes modeled as output signals of nonlinear dynamic systems in unstable environments with non-stationary, non-Gaussian interference. Traditional computational forecasting methods are often ineffective for such chaotic processes, which exhibit exponential divergence of trajectories. We propose a solution based on multidimensional correlations with other processes in the same environment. Our main hypothesis, supported by previous research, is that the dynamics of mutual connections have higher inertia than the initial processes. This allows us to form a short-term forecast using modified multidimensional regression analysis techniques.

2023 Grant 22-18-00588 articles