Influence of T-norm and T-conorm Operators in Fuzzy ID3 Algorithm

Kantarci S., KINAY A. Ö., VAHAPLAR A., Nasibov E.

IEEE International Conference on Fuzzy Systems (FUZZ-IEEE), İstanbul, Turkey, 2 - 05 August 2015 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/fuzz-ieee.2015.7337994
  • City: İstanbul
  • Country: Turkey
  • Keywords: Classification, Fuzzy Number, Fuzzy ID3 Algorithm, T-norm operators, T-conorm operators, DECISION TREES, SETS, CONNECTIVES, SYSTEMS
  • Dokuz Eylül University Affiliated: Yes


Fuzzy ID3 algorithm is a widely studied classification algorithm in order to induce fuzzy decision trees. The structure of fuzzy decision trees combines the interpretability of the decision trees with the capability of fuzzy logic on uncertainty. The fuzzy inference procedure of fuzzy decision trees is a crucial manner for the classification problems. So, the choice of the aggregation operators for the inference procedure is very important. The aim of this study is to view the performance of Fuzzy inference procedure of Fuzzy ID3 algorithm within different triangular norm (T-norm) and triangular conorm (T-conorm) operators. The fuzzy inference procedure was applied by using four different T-norm and T-conorm operators (Zadeh, Product-Sum_Umano, Yager and Hamacher). These operators were applied within the fuzzy inference phase of the system. Fuzzy c-means clustering algorithm (FCM) was used for the fuzzification of datasets. After the pre-fuzzification process, Fuzzy ID3 algorithm was performed on seven numerical datasets (Appendicitis, Balance, Iris, Wine, Hearth, New Thyroid, Haberman) selected from the KEEL data set repository. The study was encouraged with the statistical tests. The hypothesis validation techniques were used. In future studies, it is expected that by using different parameters for parametrized operators, the best accuracy rates can be handled.