Building Predictive Models with Structural Discontinuities Based on Fuzzy Markov Chains

An approach to modeling a multidimensional time series of atmospheric temperature data based on fuzzy Markov chains is considered. This approach allows us to solve the general scientific problem of constructing time series models with structural discontinuities, as well as solving the problem of parametric identification in the case when the model parameters depend on time. The paper compares the constructed models based on fuzzy Markov chains with other models proposed by the authors earlier. A numerical experiment using these data as an example showed the best quality of the proposed models, as well as a strictly justified approach to identifying areas of homogeneity of the time series.