This paper presents a Multi-Objective Evolutionary Algorithm (MOEA) for tuning type-2 fuzzy sets and selecting rules and conditions on Fuzzy Rule-Based Classification Systems (FRBCS). Before the tuning and selection process, the Rule Base is learned by means of a modified Wang-Mendel algorithm that considers type-2 fuzzy sets in the rules antecedents and in the inference mechanism. The Multi-Objective Evolutionary Algorithm used in the tuning process has three objectives. The first objective reflects the accuracy where the correct classification rate of the FRBCS is optimized. The second objective reflects the interpretability of the system regarding complexity, by means of the quantity of rules and is to be minimized through selecting rules from the initial rule base. The third objective also reflects the interpretability as a matter of complexity and models the quantity of conditions in the Rule Base. Finally, we show how the FRBCS tuned by our proposed algorithm can achieve a considerably better classification accuracy and complexity, expressed by the quantity of fuzzy rules and conditions in the RB compared with the FRBCS before the tuning process.
|Title of host publication||Advances in Fuzzy Logic and Technology 2017 - Proceedings of|
|Subtitle of host publication||EUSFLAT-2017 – The 10th Conference of the European Society for Fuzzy Logic and Technology, IWIFSGN’2017 – The 16th International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets|
|Editors||Janusz Kacprzyk, Eulalia Szmidt, Slawomir Zadrozny, Maciej Krawczak, Maciej Krawczak, Krassimir T. Atanassov|
|Number of pages||11|
|State||Published - 2018|
|Event||10th Conference of the European Society for Fuzzy Logic and Technology, EUSFLAT 2017 and 16th International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets, IWIFSGN 2017 - Warsaw, Poland|
Duration: 11 Sep 2017 → 15 Sep 2017
|Name||Advances in Intelligent Systems and Computing|
|Conference||10th Conference of the European Society for Fuzzy Logic and Technology, EUSFLAT 2017 and 16th International Workshop on Intuitionistic Fuzzy Sets and Generalized Nets, IWIFSGN 2017|
|Period||11/09/17 → 15/09/17|
Bibliographical notePublisher Copyright:
© 2018, Springer International Publishing AG.
- Fuzzy rule-based classification system
- Multi-objective evolutionary algorithm
- Type-2 fuzzy sets