Network Traffic Prediction Using Gradient Boosting Ensemble Method

6G networks will be complex heterogeneous networks. One of key challenges is how 6G mobile operators deploy and manage their networks efficiently and satisfy different requirements of mobile users. Thus, it is important to monitor the network traffics and evaluate the network performances in real time. The network traffic prediction is one of key tasks when managing the mobile networks. There are many developed network traffic models to capture the statistical characteristics of the actual network traffics. However, it is still difficult to predict network traffic accurately as the networks are getting more complex and the mobile operators should consider many different aspects such as user behaviours, traffic congestion, different network types and so on. There are many attempts to adopt AI algorithms for 6G systems because Al allows us to find network traffic patterns and adopt to varying network conditions. In order to improve the network performance and deploy the network efficiently, network traffic classification, network traffic prediction, anomaly detection and fault detection using AI are widely investigated now. In this paper, we investigate a low complexity network traffic prediction method using gradient boosting ensemble methods and evaluate its performance in a short-term network traffic period. In order to reduce the complexity and convergence time, we find the optimal number of base leaners while not degrading the performance. Under the given dataset and simulation configuration, we obtain the accurate network traffic prediction method with a low RMSE value and predict the future network traffics.