Application of Machine Learning to Predict CO₂ Emission from Transport Sector to Mitigate Climate Change

Abstract

The main objective of this chapter is to study the effect of different features of a vehicle on CO2 emissions and to create a machine learning model that can accurately predict CO₂ emissions by any vehicle. After extensive data analysis, it is concluded that features like “FUEL CONSUMPTION CITY”, “FUEL CONSUMPTION HWY”, and “VEHICLE CLASS” have a direct effect on CO₂ emissions. As the fuel consumption on the highway and in the city increases or the vehicle size increases, CO₂ emissions will also increase. On the other hand, a feature like COMB, basically the economy of the car, is inversely proportional to CO₂ emissions, i.e., better the economy, lower the CO₂ emissions. Thus, the focus should be given to increase the economy of the vehicle which will help to achieve the net-zero emissions target. The two superior models, namely the random forest regressor and XGBoost regressor, can accurately predict the carbon emissions of any kind of vehicle that run on fossil fuels. These models can be used to predict carbon emissions caused by vehicles in metropolitan cities, which in turn can help the local governing bodies in regulating the rules and public transport systems, leading to a reduction in CO₂ emissions. In future, the current dataset, which is mainly about light motor vehicles, can be extended to include the data for other types of vehicles. The combined dataset can be used to create better insights regarding the CO₂ emissions and the type of vehicle responsible for it, and hence help make necessary changes or modifications in the manufacturing of those vehicles as well as policy formulation.