Resumen
The paper deals with the problems of structural and parametric identification of cognitive models by the example of Sylov?s fuzzy cognitive maps (FCM). It is demonstrated that the problem of parametric identification can be solved using two types of methods: expert and statistical ones. An approach to the FCM parametric identification is described based on the use of methods for constructing fuzzy set adjectives: T. Saaty?s pairwise comparison method and R. Yager?s method of level sets. Problems arising when applying these methods within the context of the specified task are considered. For both methods, modifications are proposed to solve the identified problems. Also issues of building FCMs based on statistical data are discussed. For the case when data are presented in the form of spatial sampling, a method for identifying FCM parameters is proposed based on the use of multiple regression analysis. For the case when data are presented in the form of time series, a modification of this technique is proposed, which also allows solving the problem of structural identification by applying Granger causality test. Besides, an approach to the construction of FCMs under conditions of processing heterogeneous information is described, based on coordinated application of the expert and statistical methods and techniques under study. The paper presents results of experimental validation of the modified methods and the proposed techniques confirming their efficiency. In the first part of the paper, modifications of methods for constructing fuzzy set adjectives used in FCM parametric identification are described. The second part of the work is devoted to the development and research of methods for constructing FCMs based on statistical information.