Cellular Traffic Prediction using Recurrent Neural Networks
Shan Jaffry, Syed Faraz Hasan
Abstract
Autonomous network traffic prediction will be a key feature in beyond 5G networks. In the past, researchers have used statistical methods such as Auto Regressive Integrated Moving Average (ARIMA) to provide traffic prediction. However ARIMA based models fail to provide accurate predictions in highly dynamic cellular environment. Hence, researchers are exploring deep learning techniques such as Recurrent Neural Networks (RNN) and Long-Short-Term-Memory (LSTM) to develop autonomous cellular traffic prediction models.This paper proposes a LSTM based cellular traffic prediction model using real world call data record. We have compared the LSTM based prediction with ARIMA model and vanilla Feed-Forward Neural Network (FFNN). The results show that LSTM and FFNN can accurately predict cellular traffic. However, it has been found that LSTM models converged more quickly in terms of training the model for prediction.