A New Approach to the Synthesis of Fuzzy Systems from Input-Output Data
Fuzzy logic has been applied successfully to systems modeling for ages. One of its main advantages is that it provides an understandable knowledge representation. Nevertheless, most investigations have focused their efforts on achieving accurate models and by doing so, they have omitted the linguist...
| Author: | |
|---|---|
| Format: | doctoral thesis |
| Status: | Published version |
| Publication Date: | 2005 |
| Country: | España |
| Institution: | CBUC, CESCA |
| Repository: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/9147 |
| Online Access: | http://www.tdx.cat/TDX-1116105-110034 http://hdl.handle.net/10803/9147 |
| Access Level: | Open access |
| Keyword: | Lògica difusa intelligible modeling modelat intel.ligible modelado inteligible Lógica difusa Fuzzy logic Intel.ligència artificial 62 |
| Summary: | Fuzzy logic has been applied successfully to systems modeling for ages. One of its main advantages is that it provides an understandable knowledge representation. Nevertheless, most investigations have focused their efforts on achieving accurate models and by doing so, they have omitted the linguistic capabilities of fuzzy logic.<br/><br/>This thesis researches into the issues related to intelligible fuzzy models, because since science demonstrated the use of fuzzy logic when searching optimal models in terms of error (in fact a fuzzy model is a universal approximator), some but few investigators have focused their efforts in order to achieve really intelligible models in spite of losing some accuracy.<br/><br/>In this work we propose a whole methodology able to find an intelligible fuzzy model in a local manner (rule by rule) from input-output data. In this sense we find the number and position of the necessary fuzzy sets and also the linguistic rules related to them. For this purpose we have developed a hierarchical process which takes into account several steps and techniques, some of which are original contributions.<br/><br/>The resulting method is very simple and also intelligible. Therefore, this solution performs the final models with a low computational cost, but furthermore, allows the tuning of its different options depending on the nature of the problem and the characteristics of the users.<br/><br/>In this thesis we explain the whole methodology and illustrate its advantages (but also its problems) with several examples which are benchmarks in most cases. |
|---|