Delay prediction in IoT using Machine Learning Approach

Abstract

The combination of 5G, artificial intelligence, and the Internet of Things will have a great impact on future generations of wireless networks. Internet of Things (IoT) is expected to have important traffic exchange in future wireless networks. It is a generic term for technologies that allow devices to communicate with each other. These are wired and wireless sensing systems that send information from one device to another. The network traffic prediction problem includes the prediction of future network traffic characteristics from observations of past traffic. Network traffic forecasting has many applications including network monitoring, resource management, and fault detection. Machine learning (ML) has been successfully applied to traffic prediction. ML technologies have proven capable of capturing nonlinear patterns in data, making them a good candidate for traffic prediction. In this paper, we perform the delay prediction in IoT communication using a multistep ahead prediction (MSP) and single-step ahead prediction (SSP) with Time Series NARX Recurrent Neural Networks. The prediction accuracy has been evaluated using three neural network training algorithms: Trainlm, Traincgf, Trainrp, with MSE as performance function in terms of using root mean square error (RMSE) and mean absolute percentage error (MAPE) as prediction accuracy measure.