Home // International Journal On Advances in Software, volume 11, numbers 1 and 2, 2018 // View article

Fuzzy Regression Model with Modified Kernel and its Application to a Set of Real Data

Authors:
Kiyoshi Nagata
Michihiro Amagasa

Keywords: Fuzzy regression model; Kernel method; Decision making

Abstract:
Regression model is a popular and powerful model for finding a rule from large amount of collected data. It is widely used in various areas for predicting the value derived from observable values. Especially in multivariate numerical analysis, several types of regression models, not only linear but also polynomial or exponential, are established. In case of non-numerical data, although fuzzy regression models are proposed and investigated by some researchers, most of them are linear models. In order to construct a non-linear regression model with fuzzy type data set, new type of devices are needed since fuzzy numbers have a complicated behavior in multiplication and division. In this paper, we try to extend a linear fuzzy regression model to non-linear model by adapting a modified kernel method. Then, we apply the model of only low degree of polynomial kernel to a data set obtained by conducting questionnaire survey on purchasing decision making of electric assisted bicycle in Japan.

Pages: 55 to 64

Copyright: Copyright (c) to authors, 2018. Used with permission.

Publication date: June 30, 2018

Published in: journal

ISSN: 1942-2628