Machine Learning Driven Fine-Grained Spatial-Temporal Cellular User Traffic Prediction
StatisticsView Usage Statistics
Good traffic modeling and prediction are essential for providing high-quality telecommunication services, yet a challenging task. Due to the diverse network demands of Internet-based apps, the cellular traffic from an individual user can have a wide dynamic range. Most existing methods model traffic patterns as probabilistic distributions or stochastic processes and impose stringent assumptions over these models. Such assumptions may be beneficial at providing closed-form formula in evaluating prediction performance, but fall short for practice use.In this thesis, we propose STEP, a fine-grained Spatial-Temporal user traffic Prediction mechanism for cellular networks driven by machine learning algorithms. A deep graph convolution network, named GCGRN, is proposed. It is a novel combination of the graph convolution network (GCN) and gated recurrent units (GRU), which exploits graphical neural network to learn an efficient spatial-temporal model from a user’s massive dataset for traffic prediction. The prototype of STEP has been implemented and evaluated. Experimental results demonstrate that our method consistently outperforms the state-of-the-art time-series based approaches and we also show through a set of in-field tests that STEP merely imposes mild energy consumption and communication overhead to mobile devices.