Machine Learning Prediction of Multi-Band Clutter Loss using Deep Neural Networks

Efficient interference free spectrum sharing between Department of Defense (DoD) and commercial cellular systems rely on accurate prediction of clutter loss. Deployment of cellular services into a new band requires new clutter loss statistics for the band. In this paper, we assume that for a given cellular deployment, a set of band specific field validated clutter loss statistics exist. We then present an extension to a previously developed deep learning neural network (DNN) [1] to accurately predict the clutter loss statistics in bands outside of the training set. The results presented suggest that machine learning can leverage existing field validated band specific clutter loss statistics to predict the clutter loss statistics in new bands. The approach offers saving in cost, time as well as computational resources.