On Developing A New Method of Spectrum Demand Calculation for Mobile Broadband: Early Results from Auto Machine Learning Library

This study aims to report the development of a new method for spectrum demand calculation for mobile broadband by using machine learning. The new method is built based on 2 underlying concepts: the use of throughput data and the use of machine learning methods. In the current paper, we present the early results of the use of the Auto Machine Learning library PyCaret and Prophet. By comparing with statistical methods, it is found that the machine learning method has better performance. The current analyses show that both auto machine learning PyCaret and Prophet can handle data trends. Also, it is found that the monthly downlink throughput data, which is obtained from crowdsourcing, can be used as an alternative to traffic data.