Title: A new experimental design to predict carbon dioxide emissions using Boruta feature selection and hybrid support vector regression techniques

Authors: Pelin Akin; Fatih Çemrek

Addresses: Department of Statistics, Çankiri Karatekin University, Cankiri, Turkey ' Department of Statistics, Eskişehir Osmangazi University, Eskişehir, Turkey

Abstract: The problem of CO₂ emissions is a critical environmental problem for all countries. Countries want to continue to grow by striking a balance between energy and carbon emissions. This study aims to estimate the CO₂ emissions of the G7 countries with a new experimental design. This approach combines hybrid SVR with Boruta feature selection. In order to compare this model, three scenarios are built up. This new experimental approach is the first possible situation. The first scenario is this new experimental design. In the second scenario, only the hybrid SVR is implemented without feature selection. In the third scenario, only SVR was applied. The models are compared with the error terms. The best model is the first scenario with the smallest error values. G7 countries' CO₂ emissions and forecast values are close for 2019 and 2020 with the first scenario. Consequently, the CO₂ emission of G7 countries can be predicted by a model for the future period.

Keywords: hybrid support vectors; CO₂ emission; Boruta feature selection; genetic algorithm; GA.

DOI: 10.1504/IJGW.2024.136513

International Journal of Global Warming, 2024 Vol.32 No.3, pp.296 - 308

Received: 12 May 2023
Accepted: 06 Jul 2023
Published online: 05 Feb 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article