Prediction of HF Spectral Occupancy over Eastern Mediterranean Region Using Machine Learning

The ability of various machine learning (ML) algorithms to analyse and predict spectral occupancy in the high frequency (HF) spectrum is compared. One year of data (June 2012–June 2013) of HF spectral occupancy measurements collected by a dedicated measurement system installed in Cyprus is investigated. To advance the data analysis from observational to predictive, eight ML algorithms are applied one by one to model the HF spectral occupancy. The procedures for measuring and modeling congestion as a function of time of the day and day of the year are featured. Results show that XGBoost algorithm performs best with Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) as low as 0.02 and 0.01 respectively; while the correlation is as high as 0.96. Results may encourage the use of ML models in frequency usage planning and management to reduce spectral congestion in the HF broadcast bands.