PM2.5 Concentration Prediction Based on Winters’ and Fourier Analysis with Least Squares Methods in Çerkezköy district of Tekirdağ

Abstract

The rapid increase of the human population and industrialization rate in the globalizing world poses an important risk in terms of air pollution. Air pollution is an especially important issue for public health. Making the accurate predictions for air pollutants is an important step to take necessary measures. In this study, forecasting analysis for the future period was made by using the monthly average concentration values of Particulate Matter (PM2.5) causing air pollution in the Çerkezköy district of Tekirdağ province between January 2017 and April 2020. "Winters’ Method” and “Fourier Analysis with Least Squares Method” were used as the prediction approach. Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) model performance criteria were calculated based on the predictive values and actual values obtained. Whether the methods with structurally different algorithms differ in terms of prediction success was examined. Using the prediction methods, predictions for the next 20 months for PM2.5 values were obtained. The predictive values obtained from both methods were intended to create a preliminary study value for decision makers and strategists working on air pollution.