Forecasting of electromagnetic radiation time series: An empirical comparative approach

Abstract

This study compares the performance of time series models for forecasting electromagnetic radiation levels at Yesilce neighborhood in Mus, Turkey. To make successful predictions using EMF time series, which is obtained in the 36-month measurement process using the calibrated Wavecontrol SMP2 device, nine different models were used. In addition to Mean, Naive, Seasonal Naïve, Drift, STLF and TBATS standard models, more advanced ANN models such as NNETAR, MLP and ELM used in the R software environment for forecasting. In order to determine the accuracy of the models used in the EMF time series used in the study, mean absolute error (MAE), relative mean absolute error (RMAE) metrics were used. The best results obtained with NNETAR, Seasonal Naïve, MLP, STLF, TBATS, and ELM models, respectively. © ACES