Loading...
Fuzzy rules generation and extraction from support vector machine based on kernel function firing signals
Author(s)
Pitiranggon, Prasan
Benjathepanun, Nunthika
Banditvilai, Somsri
Boonjing, Veera
Date Issued
August 1, 2010
Type
Article
Abstract
Our study proposes an alternative method in building Fuzzy Rule-Based System (FRB) from Support Vector Machine (SVM). The first set of fuzzy IF-THEN rules is obtained through an equivalence of the SVM decision network and the zero-ordered Sugeno FRB type of the Adaptive Network Fuzzy Inference System (ANFIS). The second set of rules is generated by combining the first set based on strength of firing signals of support vectors using Gaussian kernel. The final set of rules is then obtained from the second set through input scatter partitioning. A distinctive advantage of our method is the guarantee that the number of final fuzzy IF- THEN rules is not more than the number of support vectors in the trained SVM. The final FRB system obtained is capable of performing classification with results comparable to its SVM counterpart, but it has an advantage over the black-boxed SVM in that it may reveal human comprehensible patterns.
Citation
World Academy of Science Engineering and Technology, 68, 239-246, 2010