Characterization of Path Loss in the VHF Band using Neural Network Modeling Technique

Abstract

Artificial Neural Networks (ANNs) which are one of the main tools used in machine learning have often been utilised in developing models for path loss modelling in recent times. However, the ANN algorithm that provides the best results has not been well established neither has the models been characterized to limit their performances and applications in the various frequency bands. In this paper, we characterize the propagation loss in the Very High Frequency Band (VHF, 30-300MHz) by using different ANN learning algorithms and activation functions based on the measurement data collected at 203.25 MHz in an urban environment (Ilorin, Nigeria). Prediction results of Hata, ECC-33, Egli and COST 231 propagation models at varying distances were fed into a feed-forward neural network and mapped to each corresponding measured path loss value. Statistical analysis shows that the ANN model that was trained with hyperbolic tangent activation function (HTAF), Levenberg-Marquardt (LM) algorithm, and 80 neurons in the hidden layer produced the most satisfactory results with Mean Error (ME), Root Mean Square Error (RMSE), Standard Deviation (SD), and coefficient of determination (R^2) values of 3.75 dB, 5.10 dB, 3.46 dB, and 0.95. However, the HTAF with Scale Conjugate Gradient (SCG) is more stable even though its prediction errors were slightly higher than that of LM.