Transportation Co2 Emission Forecasting Method Based On Ensemble Learning Approach

Abstract

In this study, a transportation CO₂ emission forecasting method is developed using the ensemble learning approach. Initially, the dataset of transportation carbon emissions was collected from the open-source database. Following that, preprocessing and data splitting were done to train and validate the performance of the machine learning (ML) algorithms. In this work, six ML algorithms were individually trained and tested for the pre-processed data. Out of these algorithms, the best two algorithms were selected for designing the ensemble learning approach. In addition, the hyper-parameter tuning of the selected ML algorithms was done using the Optuna optimizer. The simulation result indicates that the proposed method achieves lower error metric values, such as an RMSE value of 38.536, an MAE value of 26.087, and a MAPE value of 0.014, and outperforms the existing approach.