Empirical Mobile Network Traffic Prediction: Statistical Comparative Performance Analysis Using Mimo Rbfn Network Model

The exponential demand of telecommunication traffic require the development of different forecasting models in order to help industry players to plan for the future. The available forecasting traffic models are mostly developed for single-input single-output traffic data. The study applies multiple-input multiple-output (MIMO) radial basis neural network (RBFNN) model to instantaneously forecast five different time spans of telecommunication network traffic obtained from 4G and 3G networks operators. The data was taken from 3G uplink hourly, 3G daily voice, 4G weekly, 3G downlink monthly and 3G downlink quarterly from 2015 to 2017. The results prove that MIMO RBFNN (5-10-5) gives higher prediction accuracy than the other three MIMO models when subjected to different statistical tests