Examining EMF Time Series Using Prediction Algorithms With R

Abstract

In this study, electric field strength (E) levels of high-voltage lines were measured monthly in Sutluce (38 degrees N, 41 degrees E), Mus, Turkey, between 2014 and 2018, and the obtained 60 monthly mean values were used as time series (TS) to forecast next 12 months by using conventional statistical and deep learning (DL) algorithms. In addition to the conventional statistical and DL algorithms developed for R, advanced algorithms, such as long short-term memory (LSTM) and recurrent neural network (RNN) derivative, were also used in the prediction process. We applied a cross-validation technique to the electromagnetic field (EMF) data set obtained in a period of 60 months. Thus, multiple algorithm performances were compared for the same data set. In addition, the electrical field values measured in the Sutluce neighborhood were found to exceed the International Commission on Non-Ionizing Radiation Protection (ICNIRP) values.