Computing With Words in fuzzy cognitive maps

摘要

We extend our previously proposed new structure for fuzzy cognitive maps (FCM) using the weighted power mean (WPM) operator to the domain of Computing With Words (CWW), where all node activations and link strengths are described using interval type-2 fuzzy membership functions which represent words. This FCM structure is both intuitive and easier to specify than crisp FCM structures, as it allows human experts to account for the imprecision that is typical of their judgments of the network parameters. We describe a computationally feasible approach to implementing this new FCM, and provide examples illustrating its behavior, including a comparison of its converged word states with the converged states of the corresponding scalar version.