Quantifying CO2 Emissions in Vehicles Using Machine Learning

In the context of the need for environmentally friendly transportation, CO2 Emissions need to be reduced significantly to reduce the overall carbon footprint. Cars are a major source of CO2 emission. The machine enables accurate quantification and prediction of patterns in the data. In this study, the utilization of a large Canadian dataset with 7385 data to estimate CO2 emissions was used. First, there is the introduction to the dataset and its corresponding attributes, followed by defining the learning parameters and tuning them using hyperparameter optimization to obtain the optimum parameters for multiple regression models. Further, evaluation of the mean square error, mean absolute percentage error and R2 measure for evaluating models' predictive performance was performed. Finally, an evaluation of the performance and effectiveness of different machine learning models using the quantitative metrics mentioned above along with qualitative analysis from Shapley Additive Explanations and XGBoost Feature Importance analysis was performed.