PRV-FCM: An extension of fuzzy cognitive maps for prescriptive modeling

In this paper, we present a methodology based on fuzzy cognitive maps (FCMs) and metaheuristic algorithms to generate prescriptive models, called PRescriptiVe FCM (PRV-FCM). FCMs are a set of concepts interrelated that describe the behavior of a system. This kind of modeling has been extensively used to build descriptive and predictive models. We propose an extension of FCMs to develop prescriptive models and support decision-making in different domains. This adaptation characterizes FCMs, using system and prescriptive concepts. After that, it uses a metaheuristic algorithm (in this case, we use a genetic algorithm) to optimize prescriptive concepts based on system concepts and the stability of the FCM. Our proposed prescriptive approach was implemented and tested in four scenarios where it demonstrated its capability to find solutions that lead to desired values for the variables of interest. Specifically, no significant differences were found between the values of the prescriptive variables in the datasets and those generated by PRV-FCM.

Automated summary

In this paper, a methodology called PRV-FCM is presented, which combines fuzzy cognitive maps (FCMs) and metaheuristic algorithms to generate prescriptive models. The proposed approach uses a genetic algorithm to optimize the prescriptive concepts based on system concepts and the stability of the FCM. Experimental results in four scenarios demonstrate the capability of PRV-FCM to find solutions that lead to desired values for the variables of interest.