Data Processing Method for an Interval-Type Fuzzy Regression Model

Abstract

An interval model such as an interval-type fuzzy regression or an interval-type fuzzy time series illustrates the states of an analysis target as possibilities according to its interval outputs. As the possibilities of an analysis target are presented by the output of its model, irregularly distributed data distorts the shape of an interval model. In addition, the purpose of an interval model is not to make the center of distributed data and an interval model coincide, but to illustrate the possibility intervals. An interval model has these two problems, but has been improved thus far. To overcome these issues, a method that considers that a possibility grade that includes vagueness in an interval-type fuzzy regression has been proposed. This method processes samples by leveraging two different approaches. The first approach deals with vagueness in samples that improve an evaluation function, while the second approach processes other samples according to the possibility concept. Owing to the fact that the proposed method was more effective than expected, it has been verified and discussed in this paper.