5G network traffic control: a temporal analysis and forecasting of cumulative network activity using machine learning and deep learning technologies

Abstract

In fifth generation (5G), traffic forecasting is one of the target areas for research to offer better service to the users. In order to enhance the services, researchers have provided deep learning models to predict the normal traffic, but these suggested models are failing to predict the traffic load during the festivals time due to sudden changes in traffic conditions. In order to address this issue, a hybrid model is proposed which is the combination of autoregressive integrated moving average (ARIMA), convolutional neural network (CNN) and long short-term memory (LSTM), called as ARIMA-CNN-LSTM, where we forecast the cumulative network traffic over specific intervals to scale up and correctly predict the availability of 5G network resources. In the comparative analysis, the ARIMA-CNN-LSTM is evaluated with well-known existing models, namely, ARIMA, CNN and LSTM. It is observed that the proposed model outperforms the other tested deep learning models in predicting the output in both usual and unusual traffic conditions.