Machine Learning Based Mobile Traffic Forecasting for Large Scale 5G Networks

In this study, we extract and simulate traffic patterns from more than 9,000 base station cells installed in different metropolitan areas per the need that the service provider must comprehend and forecast mobile traffic in large scale cellular networks to govern and manage base station placement, load balancing, and network planning mechanisms. To accomplish this, we create, implement, and assess a method in which cells are first categorized by their point of interest and then clustered based on their temporal distribution in each region. The proposed model has been tested using real world 5G mobile traffic datasets collected over 31 weeks in various cities. Our proposed model is capable of predicting mobile traffic patterns up to 2 weeks in advance and outperforms the naive baseline method by 2.5% overall.