A K-means Interval Type-2 Fuzzy Neural Network for Medical Diagnosis

• Published in: International journal of fuzzy systems Vol. 21; no. 7; pp. 2258 - 2269
• Main Authors: Le, Tien-Loc, Huynh, Tuan-Tu, Lin, Lo-Yi, Lin, Chih-Min, Chao, Fei
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2019; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Artificial neural networks; Cardiovascular disease; Classification; Classifiers; Cluster analysis; Clustering; Computational Intelligence; Data mining; Datasets; Diagnosis; Engineering; Fuzzy logic; Heart; Machine learning; Management Science; Mathematical functions; Medical diagnosis; Neural networks; Operations Research; Pattern recognition; Stability analysis; Tumors; Vector quantization; Classification problem; Interval type-2 fuzzy neural network; Medical diagnosis; means clustering algorithm
• ISSN: 1562-2479; 2199-3211
• DOI: 10.1007/s40815-019-00730-x

This paper proposes a new medical diagnosis algorithm that uses a K -means interval type-2 fuzzy neural network (KIT2FNN). This KIT2FNN classifier uses a K -means clustering algorithm as the pre-classifier and an interval type-2 fuzzy neural network as the main classifier. Initially, the training data are classified into k groups using the K -means clustering algorithm and these data groups are then used sequentially to train the structure of the k classifiers for the interval type-2 fuzzy neural network (IT2FNN). The test data are also initially used to determine to which classifier they are best suited and then they are inputted into the corresponding main classifier for classification. The parameters for the proposed IT2FNN are updated using the steepest descent gradient approach. The Lyapunov theory is also used to verify the convergence and stability of the proposed method. The performance of the system is evaluated using several medical datasets from the University of California at Irvine (UCI). All of the experimental and comparison results are presented to demonstrate the effectiveness of the proposed medical diagnosis algorithm.