An Increased Fuzzy Min-Max Neural Network for Data Clustering

Abstract

The Fuzzy Min-max Neural Network (FMNN) is a neural network clustering based on the form of hyperboxes for classification and prediction. This paper presents an enhanced neural network model based on the fuzzy min-max clustering neural network of Simpson. The improved model which is called the Increased FMNN (IFMNN) overcomes some limitations and improves the performance of FMNN clustering. IFMNN has two main contributors to enhance the learning algorithm of FMNN. First, IFMNN adds more cases to find out overlapping In expanding hyperboxes that FMNN did not point out. Second, IFMNN gives new rules to adjust the hyperboxes contraction when finding out overlapping added in the first case. The experiments were conducted on our data set consisting of 36 patterns with two attributes and Wine data set to compare IFMNN with FMNN announced previously.