Predicting the Occupancy of the HF Amateur Service with Neural Network Ensembles

摘要

The Amateur Service is allocated approximately 3 MHz of spectrum in the HF band (3-30MHz) which is primarily used for long range communications via the ionosphere. However only a fraction of this resource is usually available due to unfavourable propagation conditions in the ionosphere imposed by solar activity on the HF channel. In this respect interference is considered a significant problem to overcome, in order to establish viable links at low transmission power. This paper presents the development of a set of Neural Network ensembles which can serve as a tool for predicting the likelihood of interference in the frequency allocations utilized by amateur users. The proposed approach successfully captures the temporal and long-term solar dependent variability of congestion, formally defined as the fraction of channels within a certain frequency allocation with signals exceeding a given threshold.