ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 3. Vol. 26. 2020

DOI: 10.17587/it.26.137-144

Yu. A. Zack, D. Sc., e-mail: yuriy_zack@hotmail.com

Fuzzy Linear Regression Analysis, Taking into Account the Influence of Input Factors

A solution is presented for fuzzy regression analysis problems in conditions where the input and output variables are represented by normalized fuzzy sets with an LR representation of the most general form membership function, and the regression coefficients are negative or positive real numbers. The free term of the regression equation is a fuzzy set of the most general form. There are restrictions on the degree of influence of some input factors established by experts. The approximation criterion is the minimum absolute value of the average unweighted sum of the absolute values of the coordinates of the minimum and maximum abscissa values of the cross sections for the membership functions of fuzzy sets of the output variable and its evaluation by the fuzzy regression model. Regression coefficients are calculated as a result of solving a certain subset of linear programming problems with the subsequent choice among them of the solution with the best value of the optimality criterion.
Keywords: fuzzy regression models, fuzzy-sets, LR-representation of the membership function, cross sections of membership functions, linear programming

P. 137–144

To the contents