Support Vector Machine for Path Loss Predictions in Urban Environment

Abstract

Path Loss (PL) propagation models are important for accurate radio network design and planning. In this paper, we propose a new radio propagation model for PL predictions in urban environment using Support Vector Machine (SVM). Field measurement campaigns are conducted in urban environment to obtain mobile network and path loss information of radio signals transmitted at 900, 1800 and 2100 MHz frequencies. SVM model is trained with field measurement data to predict path loss in urban propagation environment. Performance of SVM model is evaluated using Mean Absolute Error (MAE), Mean Square Error (MSE), Root Mean Square Error (RMSE) and Standard Error Deviation (SED). Results show that SVM achieve MAE, MSE, RMSE and SED of 7.953 dB, 99.966 dB, 9.998 dB and 9.940 dB respectively. SVM model outperforms existing empirical models (Okumura-Hata, COST 231, ECC-33 and Egli) with relatively low prediction error.