Intuitionistic fuzzy grey cognitive maps for forecasting interval-valued time series

Abstract

In many real-world forecasting problems, the time series under investigation can be approximated. In that case, instead of dealing with its exact values, only their minima and maxima achieved in the predefined periods are considered. Such an approximation forms interval-valued time series (ITS). To forecast ITS, we propose a new method that relies on fuzzy cognitive maps (FCMs). We adapt standard FCMs to the forecasting of ITS using interval-valued intuitionistic fuzzy sets. In this way, we develop a forecasting model called the Intuitionistic Fuzzy Grey Cognitive Map (IFGCM). We validate our IFGCM using publicly available stock market data for 10 indexes for which the estimation of potential investment losses (minima) and gains (maxima) is crucial. The results of these experiments prove the high efficiency of the IFGCM, especially compared with state-of-the-art models.