Iterative gradient descent approach to multiple regression with fuzzy data

Author

Bargiela, Andrzej ; Nakashima, Tomoharu ; Pedrycz, Witold

Author_Institution

Sch. of Comput. & Informatics, Nottingham Trent Univ., UK

fYear

2005

fDate

26-28 June 2005

Firstpage

304

Lastpage

309

Abstract

In this paper, we propose an iterative algorithm for multiple regression with fuzzy independent and dependent variables. While using the standard least squares criterion as a performance index we pose the regression problem as a gradient descent optimisation. Since the differentiation and summation are interchangeable we can calculate the gradient as a sum of separate components thus avoiding undue complication of analytical formulas for multiple regression. We discuss the computational complexity of the proposed algorithm.

Keywords

computational complexity; gradient methods; least squares approximations; performance index; regression analysis; computational complexity; fuzzy data; fuzzy dependent variable; fuzzy independent variable; gradient descent optimisation; iterative gradient descent; least squares criterion; multiple regression; performance index; Educational institutions; Fuzzy sets; Informatics; Iterative algorithms; Iterative methods; Least squares methods; Linear regression; Performance analysis; Regression analysis; Vectors; fuzzy data; gradient descent; multiple regression;

fLanguage

English

Publisher

ieee

Conference_Titel

Fuzzy Information Processing Society, 2005. NAFIPS 2005. Annual Meeting of the North American

Print_ISBN

0-7803-9187-X

Type

conf

DOI

10.1109/NAFIPS.2005.1548552

Filename

1548552