Evolutionary Learning of Fuzzy Rules for Regression

Abstract

The objective of this PhD Thesis is to design Genetic Fuzzy Systems (GFS) that learn Fuzzy Rule Based Systems to solve regression problems in a general manner. Particularly, the aim is to obtain models with low complexity while maintaining high precision without using expert-knowledge about the problem to be solved. This means that the GFSs have to work with raw data, that is, without any preprocessing that help the learning process to solve a particular problem. This is of particular interest, when no knowledge about the input data is available or for a first approximation to the problem. Moreover, within this objective, GFSs have to cope with large scale problems, thus the algorithms have to scale with the data.