Cluster-Based Approach for Cellular Traffic Prediction with Machine Learning Methods

Over the last decade, technologies like Fourth Generation of Cellular Communications (4G) and Cloud computing have profoundly transformed our way of living and communicating, increasing cellular traffic by tenfold every five years. Technologies like Fifth Generation of Cellular Communications (5G) and Internet of Things (IoT) are expected to continue this increasing trend. To handle such large and complex networks, the community strives to achieve self-management, intelligent Autonomous Networks. With accurate traffic prediction, we can make autonomous decisions in planning, management, resource allocation, or enforcing dynamic adaptations to the network. The goal of this paper is to design, implement, and test an efficient cellular traffic predictive system of the state of the communications network, based on selected Key Performance Indicators (KPIs), highly correlated with the cellular network traffic. This system uses a grid-based clustering algorithm to group cells according to their spatial location. We compare several time-series processing methods to forecast the Maximum Connected Users and the Download PDCP Data Volume KPIs. The persistence model is used as a baseline, the Seasonal Autoregressive Integrated Moving Average (SARIMA) is used as the classical statistical method, and the Long-Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) are used as Machine Learning methods. We show that the Recurrent Neural Network architectures (LSTM and GRU) demonstrate promising results outperforming classical statistical forecasting models significantly. The trained GRU model was successfully applied to forecast the same KPIs of other clusters (transfer learning) with similar dynamical behaviour.