A first study of Fuzzy Cognitive Maps learning using cultural algorithm

In this paper a new algorithm for Fuzzy Cognitive Maps learning is introduced. The proposed approach is based on the cultural algorithm and it is used to build the weight matrices that allow the Fuzzy Cognitive Map algorithm to find the final steady states. A Fuzzy Cognitive Map (FCM) is a fuzzy signed directed graph with feedback and models complex systems as a collection of concepts and causal relations between concepts. An FCM can be constructed by using experts' knowledge or historical data. In this paper we have developed an automated method FCM learning which uses a type of evolutionary algorithm known as a cultural algorithm. We explain the algorithm and demonstrate its performance advantages.

I. Introduction

A Fuzzy Cognitive Map (FCM) is a fuzzy weighted network used to model and analyze systems by presenting the causal relationships among system's components [7]. This technique has four advantages in analyzing complex systems: (a) it models complex problems in a simple and understandable way, (b) it is flexible in system analysis and design and has the ability to deal with fuzzy information, (c) it can represent all possible causal relationships between components of a given problem and (d) FCM inferences are carried out by numerical calculations instead of symbolic deductions [3], [4], [11].