Forecasting of Fuzzy Time Series Based on the Concept of the Nearest Fuzzy Sets and Tensor Models of Time Series

Forecasting of fuzzy time series is considered by presenting a standard fuzzy set in the form of a tensor obtained as a result of the tensor product of components, as well as by forming a tensor sequence whose last element (the predicted fuzzy set) is calculated as an incomplete tensor (a tensor with missing elements). The singular value decomposition of the restored tensor allows us to obtain a subset of ordered pairs that is the closest (in terms of the F-norm) to the predicted fuzzy set. An example of predicting a fuzzy time series is given.