A network traffic flow prediction with deep learning approach for large-scale metropolitan area network

Abstract

Accurate and timely internet traffic information is important for many applications, such as bandwidth allocation, anomaly detection, congestion control and admission control. Over the last few years, internet flow data have been exploding, and we have truly entered the era of big data. Existing traffic flow prediction methods mainly use simple traffic prediction models and are still unsatisfying for many real-world applications. This situation inspires us to rethink the internet traffic flow prediction problem based on deep architecture models with big traffic data. In this paper, we propose a novel deep-learning-based internet traffic flow prediction method, which is called SDAPM. It consider the spatial and temporal correlations inherently and internet flow data character. A stacked denoising autoencoder prediction model (SDA) is used to learn generic internet traffic flow features, and it is trained in a greedy layer-wise fashion. Moreover, experiments demonstrate that the SDAPM for traffic flow prediction has effective performance. Our prediction model is in production as part of the traffic scheduling system at China Unicom, one of the largest Internet companies in China, helping improving the network bandwidth utilization.