Prediksi Jumlah Emisi Karbon Dioksida (CO₂) Menggunakan Metode Extreme Gradient Boosting (XGBoost)

Abstract

Global climate change caused by the surge in CO₂ emissions has become a serious issue in Asia, mainly due to deforestation, the use of fossil fuels, and activities in other sectors. As a form of commitment, a number of countries in Asia have agreed to global agreements such as the Paris Agreement to reduce greenhouse gas emissions, including through the use of renewable energy, energy efficiency, and controlling emissions from various sectors such as industry, transportation, and deforestation. To support these efforts, this study develops a CO₂ emission prediction model to assist governments and stakeholders in identifying trends, evaluating policies, and formulating more optimal mitigation strategies. The method used in this study is Extreme Gradient Boosting (XGBoost). The data used are annual data from 1965 to 2023, covering Gross Domestic Product, energy consumption, fossil fuel consumption, renewable energy consumption, electricity generation, forest area, and agricultural land area in several developing countries in Asia. These countries include the Philippines, Indonesia, Malaysia, Thailand, Vietnam, India, Pakistan, Bangladesh, China, Iran, Saudi Arabia, Kazakhstan, and Uzbekistan. The data were processed through preprocessing, training using the XGBoost model, and testing to produce a prediction model with hyperparameters optimized based on experimental results. The predictions yielded evaluation scores of MSE at 1.1728, RMSE at 1.0830, MAE at 0.7530, and R² at 0.9660.