Interpretability Issues in Fuzzy Genetics-Based Machine Learning for Linguistic Modelling

Abstract

This chapter discusses several issues related to the design of linguistic models with high interpretability using fuzzy genetics-based machine learning (GBML) algorithms. We assume that a set of linguistic terms has been given for each variable. Thus our modelling task is to find a small number of fuzzy rules from possible combinations of the given linguistic terms. First we formulate a threeobjective optimization problem, which simultaneously minimizes the total squared error, the number of fuzzy rules, and the total rule length. Next we show how fuzzy GBML algorithms can be applied to our problem in the framework of multi-objective optimization as well as single-objective optimization. Then we point out a possibility that misleading fuzzy rules can be generated when general and specific fuzzy rules are simultaneously used in a single linguistic model. Finally we show that non-standard inclusion-based fuzzy reasoning removes such an undesirable possibility.

DOI: 10.1007/978-3-540-39906-3_11

Extracted Key Phrases

6 Figures and Tables

Statistics

0204060'04'05'06'07'08'09'10'11'12'13'14'15'16'17
Citations per Year

227 Citations

Semantic Scholar estimates that this publication has 227 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@inproceedings{Ishibuchi2003InterpretabilityII, title={Interpretability Issues in Fuzzy Genetics-Based Machine Learning for Linguistic Modelling}, author={Hisao Ishibuchi and Takashi Yamamoto}, booktitle={Modelling with Words}, year={2003} }