Data classification using logarithmic spiral method based on RBF classifiers

Abstract

Clustering is the organization of a set of data in homogeneous classes. It aims to classify the representation of the initial data. The automatic classification recovers all the methods allowing the automatic construction of such groups. This paper describes how to classify data using a new design of neural classifiers with radial basis function (RBF) based on a new algorithm for characterizing the hidden layer structure. This algorithm, called k-means Euclidean distance, groups the training data class by class in order to calculate the optimal number of clusters of the hidden layer, using the Mean Squared Error. To initialize the initial clusters of k-means algorithm, we have used the method of logarithmic spiral golden angle. Two examples of data sets are considered to improve the efficiency of the proposed approach and the obtained results are compared with basic literature classifier.