Affiliations: [a] Department of Electronics and Communication Engineering, Universiti Tenaga Nasional, Selangor, Malaysia | [b] School of Electrical and Computer Engineering, Xiamen University Malaysia, Selangor, Malaysia | [c] Department of Mechanical Engineering, University of Malaya, Kuala Lumpur, Malaysia
Corresponding authors: Shen Yuong Wong, School of Electrical and Computer Engineering, Xiamen University Malaysia, Selangor, Malaysia. E-mail: [email protected]. Keem Siah Yap, Department of Electronics and Communication Engineering, Universiti Tenaga Nasional, Selangor, Malaysia. E-mail: [email protected]
Abstract: The hybrid of artificial neural network (ANN) and fuzzy logic system (FLS) can expend itself dynamically in a strong discovery of explicit knowledge to solve classification and regression problems with new input patterns. In this paper, a hybrid of Generalized Adaptive Resonance Theory (GART) and interval type-2 fuzzy logic system (IT2FLS) algorithm is proposed, and named as Generalized Adaptive Resonance Theory and interval type-2 fuzzy logic system (GART-IT2FLS). The GART is a combination of adaptive resonance theory network (ART) and Generalized Regression Neural Network (GRNN). GART is capable to deal with classification task effectively. However, type-2 fuzzy sets (T2 FS) is used to represent and model the uncertainties on inputs. The performance evaluation of GART-IT2FLS algorithm in three medical datasets has proven that GART-IT2FLS is capable to learn incrementally without plasticity-stability dilemma, and model uncertainties in medical datasets. The inferences of GAR-IT2FLS in these applications are discussed. The performance results show that GART-IT2FLS has obtained a comparable number of rules. The Wisconsin Breast Cancer and Heart Disease experiments demonstrated GART-IT2FLS has improved the testing accuracies.
Keywords: Generalized adaptive resonance theory, interval type 2 fuzzy logic system, classification, medical diagnosis