Predicting carbon dioxide emissions until 2030 using conventional forecast techniques and machine learning models

  • Pedro Albuquerque (Student)

Student thesis: Master's Thesis


This dissertation investigates global carbon dioxide emissions and provides forecasts through 2030 using both conventional forecasting techniques and machine learning methods. Data on emissions by fuel type and renewable sources will be analysed for all continents and Portugal. We find that the Random Forest machine learning model produces more accurate forecasts than the conventional ARIMA. Our findings show that the predicted levels of carbon dioxide emissions on the various continents vary significantly. Emissions are predicted to rise in Asia, South America, and Africa while falling in Europe, North America, Australia, and Portugal. However, none of the continents are on track to reach the 55% reduction target set by the Paris Agreement. North America and Europe are making progress, but still 20% away from the expected values.
Date of Award3 May 2023
Original languageEnglish
Awarding Institution
  • Universidade Católica Portuguesa
SupervisorNicolò Bertani (Supervisor)


  • Carbon dioxide
  • Paris agreement
  • Machine learning


  • Mestrado em Análise de Dados para Gestão

Cite this