Akademik Veri Yönetim Sistemi
SOFT COMPUTING, cilt.2025, ss.1-32, 2025 (SCI-Expanded)
Time series analysis is a specific way of analyzing the time series data to understand how to change over time and to collect some projections about how it will change in the future. Forecasting a time variable is possible with the help of models created by observing the movements of the variable in the past. In this study, Electromagnetic Radiation (EMR) is being analyzed. Human beings are exposed to more and more EMR day by day due to the increasing number of base stations. Although the negative impacts of devices that cause on living things are known, it is impossible to remove them altogether from our lives. In this sense, it is crucial to observe, model, and forecast the amount of EMR in an environment. For this purpose, in this study, two hybrid models, ANN and ANFIS based on the EMD of Time Series were applied to the EMR measurement data: Experimental Mode Decomposition-Artificial Neural Networks (EMD-ANN) and Experimental Mode Decomposition-Adaptive Neuro Fuzzy Inference System (EMD-ANFIS). Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Normalized Root Mean Square Error (NRMSE), and Scatter Index (SI) as evaluation criteria were used to compare the model performances. In addition, for each model, the scatter plot of the forecasts against actual values with the regression line and the coefficient of determination (R2) and Nash–Sutcliffe efficiency (NSE) as a goodness of fit measure were provided to evaluate the best model. As a result of the calculations, the RMSE, MAE, MAPE, SI, Accuracy (1-NRMSE), R2, and NSE values for the test set were found to be 0.8177, 0.6372, 8.92%, 0.1159, 88.94%, 0.7313 and 0.7299, respectively, when the ANN model was used, while these values were found to be 0.3943, 0.2935, 4.09%, 0.0559, 94.67%, 0.9392 and 0.9372, respectively, in the EMD-ANN model. When the ANFIS model was used, the relevant values were found to be 0.8270, 0.6431, 8.95%, 0.1173, 88.81%, 0.7257, and 0.7237, respectively, while in the EMD-ANFIS model, these values were found to be 0.3956, 0.3045, 4.21%, 0.0561, 94.65%, 0.9369 and 0.9368, respectively. As a result, it was observed that the ANN and ANFIS methods hybridized with the EMD method gave better results than when they were used alone.