Estimation of Received Signal Power for 5G-Railway Communication Systems

This paper presents the estimation of received power signal based on Support Vector Regression (SVR). The simulated datasets are used, which contain the positions of transmitter (Tx) and receiver (Rx), the distance of TX and RX, and corresponding path loss, and the carrier frequencies. SVR presents the accuracy estimation of simulated datasets computing which shows Mean Square Error (MSE) as an average value of estimation errors that are squared, Root Mean Square Error (RMSE) as another parameter for measuring the accuracy of a estimation as a root value of MSE Average Root also R² as the coefficient of determination tool for measuring the ability of a model in explaining dependent variable variations. If the value of R² approaches one, it means that predictive results can follow variable patterns or variations well dependent. Cross Validation is a performance measurement. The aim is to find the best hyper-parameter combination so that machine learning can predict data accurately and prevent over-fitting problems. Optimal parameter values are determined by using the Grid Search Method, where machine learning will do modeling using the range C ɤ and ɛ given. Therefore, SVR Hyper-Parameter shows the most optimized parameter with C which affects the penalty given when there is an error in classification, Gamma that affects the pace of learning process, Epsilon indicates the error limit than can be ignored. The parameter values that produce the highest accuracy or the smallest error will be chosen as the best parameter.