A fuzzy cooperative genetic algorithm for fuzzy modeling
Fuzzy modeling refers to the process of identifying the fuzzy parameters by defining fuzzy rules and fuzzy sets. The process of identifying the fuzzy parameters becomes complicated and difficult to evaluate particularly when it involves complex problems in engineering and medical applications. This...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2011
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/32804/1/MohdarfianIsmailMFSKSM2011.pdf http://eprints.utm.my/id/eprint/32804/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Fuzzy modeling refers to the process of identifying the fuzzy parameters by defining fuzzy rules and fuzzy sets. The process of identifying the fuzzy parameters becomes complicated and difficult to evaluate particularly when it involves complex problems in engineering and medical applications. This is due to the fact that there are many fuzzy parameters to be identified such as the characteristics of the fuzzy rules and fuzzy sets. Besides that, the accuracy and the ability to interpret the fuzzy rules and the fuzzy sets would also need to be considered. This study proposes an improved method of fuzzy modeling called the Fuzzy Cooperative Genetic Algorithm (FCoGA) which integrates the Genetic Algorithm (GA) and Cooperative Coevolution Algorithm (CooCEA) that automatically generate and refine the fuzzy rules and the fuzzy sets. The GA is used to exploit the chromosomes that represent the fuzzy parameters whereas the CooCEA is applied to reduce the complexity of the fuzzy parameters representation which is carried out by subdividing chromosomes into three sub-chromosomes known as species. The FCoGA comprises of three phases which are simplification, tuning and evaluation. Simplification phase involves decomposition of the chromosomes into three species of chromosomes that represent the fuzzy parameters consisting of fuzzy rules, membership functions and length of the overlapping membership functions. The tuning phase involves the process of altering and tuning all the species. Lastly, the evaluation phase validates the performance of the FCoGA. Three benchmark datasets; breast cancer, diabetes and Iris have been used to evaluate the performance of the FCoGA. The experimental results showed that the FCoGA obtained the highest percentage of accuracy classification compared to other techniques such as conventional GA, multi-objective CooCEA, rule extraction and decision tree. The results also indicated that the FCoGA produced higher interpretability of a fuzzy model. |
|---|