Fuzzy Clustering Algorithm for Interval Data Based on Feedback RBF Neural Network

Abstract

Data set with missing attribute is often encountered in practical applications. To solve the problem that fuzzy c-means clustering algorithm can’t be directly used for fuzzy clustering of incomplete data, a feedback Radial Basis Function neural network (FRBF) is proposed to estimate the missing attribute values for incomplete data. The error between the actual output value of RBF neural network and the expected value is fed back to the input layer, then a feedback RBF neural network is constructed. Further, due to the numerical data can’t accurately describe the incomplete data, we provide an interval approach, which can convert the numerical data set into an interval valued data set. Thus, an interval fuzzy c-means clustering algorithm based on improved RBF neural network (FRBF-IFCM) is proposed to perform clustering analysis. Experimental results show that this algorithm has better accuracy in data clustering performance than similar algorithms.