Estimation of highway CO2 emissions by artificial intelligence methods

Global warming has been one of the biggest problems of the world. Greenhouse gases that cause global warming are increasing with increasing population and urbanization, energy production and consumption. CO₂ emissions released by the combustion of fossil fuels used in road transport significantly affect air pollution. Road transport has been constituted 94.9% of CO₂ emissions in Turkey. For this reason, the policies developed for road transport are gaining importance. Therefore, Ridge Regression, Artificial Intelligence Methods have been used for the estimation of carbon dioxide emissions from road transport in this study. The Artificial Intelligence methods used are, Artificial Neural Networks (ANN), Adaptive Neural-Fuzzy Inference Systems (ANFIS) and Fuzzy Simple Membership Functions and Rule Making Technique (Fuzzy SMRGT) methods. In this study, 31 years of data recevied from TUIK and KGM for the years 1990-2020, including the sum of all provinces, have been used for road transport emission estimation. For road transport emissions estimations; Vehicle Kilometers (VK), Passenger Kilometers (PK), Ton Kilometers (TNK), Highway Length (HL), Population Number (PL), Number of Motor Vehicles (NMV), Minimum Wage (MW), Total Energy Consumed on Highways (TECH) variables are taken into consideration as independent variables. For the Artificial Intelligence Methods, %70 of the data are divided into training, %30 of them are divided into testing. The model performances have been compared according to Adjusted Coefficient of Determination (R2adjusted), Square Root of Standardized Mean Error Squares (RMSEstandardized) and Mean Absolute Error (MAE) metrics. When the model estimates are compared; The ANN method achieved the best results among the other methods have been used, R2adjusted (0,999), MAE (0,25*103) ve RMSEstandardized (0,01) values. As a result of the results, the methods used are recommended to be used in highway CO2 prediction models.