Classification based on Gaussian-kernel Support Vector Machine with Adaptive Fuzzy Inference System

In this paper, we propose a new classification approach which combines the advantages of both Gaussian-kernel Support Vector Machine and Adaptive Fuzzy Inference System. Instead of generating a large number of candidate rules as in fuzzy classification, the proposed method adopts the decision trees to generate rules directly from training data. Decision trees provide architecture to generate fuzzy IF–THEN rules from the training data where the fuzzy parameters of the rules would be optimized using Genetic Algorithm. The Gaussian-kernel SVM will be used in the classification phase using the parameters obtained from Particle Swarm Optimization. Experimental results of the proposed approach has proved significantly better accuracy than other state-of-the-art classification methods by testing it on benchmark UCI datasets. Streszczenie. Zaproponowano nową metodę klasyfikacji łączącą zalety metod: Gaussian-Kernel Support Vector Machine i Adaptive Fuzzy Interference System. Wykorzystano drzewo decyzyjne do tworzenia zasad klasyfikacji bezpośrednio z danych treningowych. Parametry logiki rozmytej określano wykorxzystując algorytm genetyczny. A parametry SVM wykorzystując lagorytm mrówkowy. Metoda klasyfikacji bazująca na SVM I adaptacyjnej logice rozmytej