A New Hybrid Approach of Gravitational Search Algorithm with Spiral-Shaped Mechanism-based RBF Neural Network

Abstract

This article proposes a neural network and a non-linear time series method via a prediction model based on an RBF neural network. The proposed model predicts and identifies a non-linear system using the Hybrid Gravitational Search Algorithm (HGSA). The proposed algorithm HGSA is deemed with the optimal parameter settings and network topology of a neural network. GSA is implemented with a spiral-shaped mechanism (SSM) to eradicate primary drawbacks such as slow convergence. Thus, it tends to premature convergence. Moreover, HGSA-SSM selects updated particles' locations through the most suitable selection law that provides an exact match in global and local search competencies. Additionally, HGSA-SSM could optimize the RBF neural network's parameters such that a network model is generated with high precision. Hence, our proposed novel proposed model (HGSA-SSM –RBFNN) overcomes the non-linear problems by developing several numerical precedents, and it is found efficient than the existing RBF neural networks.