Novel approach for classification and prediction of non linear chaotic databases

Abstract

This paper we propose a methodology of classification and prediction of chaotic time series datasets by the beauty of particle swarm optimization (PSO) in multilayer feed forward backpropagation neural network (MFFBPNN) for finding initial weights and biases of the MLFFBPNN. Designed algorithm was used for different horizons in classification and prediction to chaotic datasets by overcoming the disadvantage of feedforward back propagation of getting stuck at local minima or local maxima. The Chaotic datasets such as Mackey glass series, Box Jerkins Gas furnace, Breast Cancer, Diabetic dataset are considered in this paper for testing of designed algorithm. The performance of PSO with MLFFBPNN for finding weights and biases is implemented and compared with random initialization of weights and biases with normal MFFBPNN. The result shows the Comparative performance of MLFFBPNN and designed algorithm on dynamic behavior of chaotic time series datasets in terms of mean square error.