Deep Learning-Based Network Traffic Prediction for IoT Routing Optimization in Communication Systems

Abstract

The fast development of the Internet of Things (IoT) has resulted in more and more complex and dynamic network traffic, which poses a serious challenge to the effective routing of the network in a resource-limited environment. This paper suggests a Variational Autoencoder (VAE)-gated recurrent unit (GRU)-based network traffic forecasting and routing enhancement framework to the IoT communication systems. The framework uses VAE to learn latent representations of network traffic, which includes latent patterns and variability, and a GRU to learn time-based dependencies so as to make accurate multi-step predictions of traffic. Traffic information is also predicted and included in a traffic-aware routing engine whereby optimal routes are chosen dynamically in response to congestion, remaining node energy, and Quality-Of-Service (QoS) needs. The Python network simulator is used in the implementation and evaluation of the system and its performance is compared with the performance of the conventional routing protocols like RPL and AODV. The simulation findings prove that the proposed VAE-GRU framework can drastically enhance the performance of the network with a Mean Absolute Error (MAE) of 0.052 and Root Mean Square Error (RMSE) of 0.074 in traffic prediction. At the network level, the predictive routing mechanism reduces end-to-end delay (61-132ms), increases packet delivery ratio (91.8-99%), enhances throughput (346-468kbps), and lowers average energy consumption per node (1.41-2.54J) across varying traffic loads. These findings substantiate the fact that predictive intelligence in traffic and adaptive routing help to provide a scalable, energy-efficient, and resilient solution to IoT networks. This framework is especially efficient in the high-traffic and dense network environment, which outlines its usability in the next-generation IoT communication systems that demand a high-quality and reliable, proactive and resource-conscious routing.