Time Series Processing with Cognitive Maps. the Case of General Forecast Modeling for Time Series of Similar Nature.

This paper aims to provide an overview of interpretation issues of Fuzzy Cognitive Maps trained to model time series. We analyze the impact of different Fuzzy Cognitive Maps design scenarios on modeling quality. In particular, our goal is to analyze the quality of modeling provided by Fuzzy Cognitive Maps with two aspects: firstly, the varied configuration of objective functions and error functions and, secondly, generalization of modeling from particular elements (time series) to the class of similar nature elements. We compare qualitatively and quantitatively time series models based on good Fuzzy Cognitive Maps designs with the aforementioned design options.