Forecasting CO2 Emissions in Industrial Sector Using Time Series and Machine Learning Modeling Methods

Abstract

This study conducts a comparative analysis of various algorithms to forecast CO2 emissions in Thailand's industrial sector using monthly data from 1987 to 2022. We investigate the performance of two modeling methods: time series analysis and machine learning. The time series analysis utilizes ARIMA, which is a traditional statistical method, and modern deep learning algorithm called Long Short-Term Memory (LSTM) network. The adopted machine learning techniques include Random Forest, Linear Regression (LR), and Support Vector Regression. The primary objective is to determine the most effective model for accurately predicting CO2 emissions over time. The data preprocessing steps, model training process, and evaluation metrics are carefully designed to ensure a fair and robust comparison. Our results demonstrate significant differences in the predictive capabilities of these models. LSTM shows superior performance in capturing complex patterns in the time series data, whereas LR and ARIMA come in second and third, respectively. This study provides valuable insights into the strengths and limitations of each approach, offering a reference point for future research and policy-making aimed at mitigating industrial CO2 emissions in Thailand.