Evaluation of Single-Input Single-Output Radial Basis Function Neural Network in Modelling Empirical 4G Traffic

Abstract

The fast improvement in the telecommunication sector in general and the accompanying revolution in the mobile communication sector provide a challenging task to telecommunication vendors and operators in satisfying the huge subscribers. To address this challenge, appropriate forecasting models must be provided. This study has developed fourteen different single input single output (SISO) Radial basis function neural network (RBFNN) with Levenberg-Marquardt (LM) and Resilient backpropagation (Rprop) algorithms. Two models, Rprop-SISO RBFNN (1-20-1) and LM-SISO RBFNN (1-20-1) were selected based on computational time and minimum values of prediction. When the performance of the models were compared with empirical 4G hourly traffic data collected from an operator in Ghana, Rprop-SISO RBFNN (1-20-1) gave a better computational time while LM-SISO RBFNN (1-20-1) results indicate minimal values of MSE and NRMSE. Therefore LM-SISO RBFNN (1-20-1) model was selected as the appropriate model to predict 4G hourly traffic.