Amin Ruhul, Ar Salan Md Sifat, Hossain Md Moyazzem
Department of Statistics and Data Science, Jahangirnagar University, Savar, Dhaka, 1342, Bangladesh.
Heliyon. 2024 Feb 6;10(4):e25416. doi: 10.1016/j.heliyon.2024.e25416. eCollection 2024 Feb 29.
The indicators of economic and sustainable development ultimately significantly depend on carbon dioxide (CO) emissions in every country. In Bangladesh, there is an increasing trend in population, industrialization, as well as electricity demand generated from different sources, ultimately increasing CO emissions. This study explores the relationship between CO emissions and other significant relevant indicators. Moreover, the authors aimed to identify which model is effective at predicting CO emissions and assess the accuracy of the prediction of different models. The secondary data from 1971 to 2020, was collected from the World Bank and the Bangladesh Road Transport Authority's publicly accessible website. The generalized additive model (GAM), the polynomial regression (PR), and multiple linear regression (MLR) were used for modeling CO emissions. The model performance is evaluated using the Bayesian information criterion (BIC), Akaike information criterion (AIC), Root mean square error (RMSE), R-square, and mean square error (MSE). Results revealed that there are few multicollinearity problems in the datasets and exhibit a nonlinear relationship among CO emissions. Among the models considered in this study, the GAM model has the lowest value of RMSE = 0.008, MSE = 0.000063, AIC = -303.21, BIC = -266.64 and the highest value of R-squared = 0.996 compared to the MLR and PR models, suggesting the most appropriate model in predicting CO emissions in Bangladesh. Findings revealed that the total CO emissions and other relevant risk factors is non-linear. The study suggests that the Generalized additive model regression technique can be used as an effective tool for predicting CO emissions in Bangladesh. The authors believed that the findings would be helpful to policymakers in designing effective strategies in the areas of a low-carbon economy, encouraging the use of renewable energy sources, and focusing on technological advancement that reduces CO emissions and ensures a sustainable environment in Bangladesh.
每个国家经济与可持续发展的指标最终都极大地依赖于二氧化碳（CO）排放。在孟加拉国，人口、工业化以及不同来源产生的电力需求都呈上升趋势，最终导致CO排放量增加。本研究探讨了CO排放与其他重要相关指标之间的关系。此外，作者旨在确定哪种模型在预测CO排放方面有效，并评估不同模型预测的准确性。1971年至2020年的二手数据来自世界银行以及孟加拉国道路运输管理局的公开网站。广义相加模型（GAM）、多项式回归（PR）和多元线性回归（MLR）被用于对CO排放进行建模。使用贝叶斯信息准则（BIC）、赤池信息准则（AIC）、均方根误差（RMSE）、决定系数（R平方）和均方误差（MSE）来评估模型性能。结果显示，数据集中存在一些多重共线性问题，并且CO排放之间呈现非线性关系。在本研究考虑的模型中，与MLR和PR模型相比，GAM模型的RMSE值最低，为0.008，MSE值为0.000063，AIC值为 -303.21，BIC值为 -266.64，决定系数（R平方）值最高，为0.996，表明它是预测孟加拉国CO排放的最合适模型。研究结果表明，CO排放总量与其他相关风险因素呈非线性关系。该研究表明，广义相加模型回归技术可作为预测孟加拉国CO排放的有效工具。作者认为，这些研究结果将有助于政策制定者在低碳经济领域设计有效策略，鼓励使用可再生能源，并专注于减少CO排放并确保孟加拉国可持续环境的技术进步。
