Machine Learning-based Frequency Resource Demand Prediction for a Mobile Satellite Network

This paper details the architecture and employment of a link-layer user-uplink simulator and a methodology that enables prediction of the spectrum demand of various connection service mixtures through the use of random forest machine learning. The results show that a model can be produced using this methodology. Predictions are used as an input to a channel load balancing algorithm, with the goal of learning the behaviour of the scheduling and framing algorithms.