Aplikace modelů s proměnlivými režimy při předvídání finančních časových řad

This work primarily deals with the creation of models for the analysis and prediction of time series of financial nature. For the analysis of financial assets during the period from 21.07.2020 to 21.07.2022, selected classes of models are described and constructed. Specifically, these include the linear AR(p) model, and nonlinear models with two regimes determined by observable variables – SETAR, LSTAR, ESTAR, as well as models with two regimes determined by unobservable variables – MSW. Subsequently, the most suitable model is selected based on the Akaike information criterion, and an estimation of future development for a period of one year, until 21.07.2023, is performed. In the conclusion of the work, the prediction accuracy is compared with actual values, and potential real-world investment applications are suggested.