Using artificial neural network approach to predict rain attenuation on Earth-space path

Abstract

A rain attenuation model based on artificial neural network is proposed in this paper without making many assumptions as traditional methods and therefore improves the prediction accuracy. Based on an analysis of various factors affecting rain attenuation, a rain attenuation model with artificial neural network is developed after training and verifying many different neural network topologies. The prediction results of the proposed model is also compared with that of the CCIR model. The results show that applying the artificial neural network to predict rain attenuation of a high frequency wave is a good approach and decreases the mean prediction error by 0.59 dB and the RMS error by 0.69 dB. The paper shows that this model is a new and effective way to predict rain attenuation with an artificial neural network.