Thailand Carbon Dioxide Emissions Forecasting using Stacked LSTM-based Prediction Model

Over the past few decades, Thailand has emerged as a prominent producer of carbon dioxide (CO2), ranking third in emissions after Poland and Taiwan. The emission of CO2 is a significant factor in global warming, which has had substantial adverse effects on both the economy and human life. Accurately predicting CO2 emissions is crucial for evaluating Thailand's existing low-carbon environmental protection policies. However, conventional forecasting models primarily rely on historical data regularities and rarely incorporate new information on development trends. "Stacked Long Short-Term Memory (LSTM) is used as an algorithm for forecasting and analyzing CO2 emissions in Thailand. The study employs a dataset covering the period from 1987 to 2021. The LSTM model is evaluated against VAR (Vector Autoregression) and ARIMA (Auto Regressive Integrated Moving Average) to assess its predictive performance. The LSTM model demonstrates significantly higher accuracy in forecasting CO2 emissions, outperforming other models with lower RMSE (698.94), MAE (609.86), and MAPE (4.54%) values, making it the most suitable choice for this study. The prediction results are deployed and visualized on a website-based application using TensorFlow.js. Furthermore, this research contributes to the advancement and expansion of LSTM theory.