time series, fuzzy transform, evaluative linguistic expressions, fuzzy set

In this paper, we address the problem of automatic recognition of structural breaks in time series. The former are unexpected shifts of the course or sudden change of the volatility of time series. Structural breaks can be caused, e.g., by changes in the organization of a company, global or local economic development, global shifts in capital and labor, various kinds of outer influences such as discovery or depletion of natural resources, etc. Structural breaks in time series are usually detected using statistical methods. In this paper, we suggest using special non-statistical techniques of fuzzy modeling. We will employ two classes of them, namely the fuzzy transform and selected methods of Fuzzy Natural Logic. The fuzzy transform enables us to estimate the average slope of time series in an area characterized by a fuzzy set. The slope is then evaluated by evaluative linguistic expressions, which enables us to identify intervals with monotonous behavior and, consequently, identify structural breaks. Our method is simple, transparent, and computationally effective.