Русский
!

Conference publications

Abstracts

XVII conference

Fuzzy cognitive maps in Artificial Intelligence

Воробьев Г.Г.

University 'Dubna', Russia, 141980, Moscow region, Dubna, University street, 19, e-mail: gvorobjev@gmail.com

1 pp. (accepted)

While analyzing a complicated situation consisting of a large number of variables, it’s difficult to estimate a total effect of all other variables on the variable that an expert is interested in. The problems of forecast and strategy search require a complete set of variables to be taken in consideration, and this could hardly be done by an expert. These problems may be solved with help of cognitive modeling.

A cognitive map is a directed graph capable of modeling interrelationships or causalities existing amongst concepts (nodes).

Fuzzy cognitive map (FCM) is a logical evolution of a cognitive map, that uses fuzzy-graph approach. FCM is formally described as a matrix [wij] of relationships between concepts, -1 ≤ wij ≤ 1. Using transitive closure, a new matrix that describes indirect relationships between all concepts can be calculated.

The methods of cognitive modeling have been implemented in a program that allows to input sentences like “increasing A causes B to strongly increase; increasing C causes A to slightly decrease”, and so on. A program user gets a detailed forecast of situation’s future, after some initial actions are performed (“what will happen, if C will vastly increase?”). He can also find out a strategy that achieves his goal state (“what should be done to slightly increase A?”). An efficiency of a program is demonstrated on a FCM “Oil price fluctuations in the world’s economy”.



© 2004 Designed by Lyceum of Informational Technologies №1533