Multi-objective Coordination Of Genetic Programming And Its Application In Fuzzy Modeling

Fuzzy Rule-Based System (FRBS) can be divided into two kinds: approximate anddescriptive. The former is mainly concerned with accuracy and flexibility. The latter isfocused on interpretability. The trade-off between accuracy and interpretability is animportant research topic. It is also discussed in this thesis.Genetic programming (GP) is a new technology for optimization, which simulatesinherit and evolution in the nature and gets optimal solutions through reproduction,crossover and mutation operations. This thesis focuses on the application of GP to modelfuzzy system. The main research contribution of the thesis can be summarized as follows:1,This thesis introduced a grammar for deriving Fuzzy Rule Bases and combined agenetic programming with a context-free language to evolve classifier systems. Themethod was called Fuzzy GP. A Fuzzy Rule Based Classifier can be described in terms of aBackus-Naur Form (BNF) grammar. 2,Against the deficiency of the fuzzy system identified with grid partition, in thisthesis a multi-objective cooperative co-evolutionary system which is based the fastnon-dominance sort is proposed for discovering fuzzy classification rules. The algorithmuses two evolutionary algorithms: a fuzzy genetic programming evolving a population offuzzy sets and a simple evolutionary algorithm evolving a population of membershipfunction definitions. Firstly of all, feature selection is accomplished by the Simba featureselection algorithms. In the end of each running, a set of Pareto optimal solutions isobtained. The proposed approach was applied to two benchmark problems, and the resultsshow its validity.