Forecasting short term European CO2 returns using high frequency gas data: a machine learning approach

Details

Supervisors

Iania, Leonardo

Faculty

Louvain School of Management

Degree label

Master [120] : ingénieur de gestion, à finalité spécialisée

Abstract

In this Master thesis, machine learning algorithms are exploited to forecast weekly returns of the European Trading System during phases III and IV, and study the impact of physical gas flows on the forecasts. Additionally, a popular dimension reduction technique called Principal Component Analysis is used together with the Marchenko-Pastur theorem. The findings are twofold: 1) information about the gas flows do not improve the accuracy of the forecasts during the phase III, but tend to improve them during the phase IV, 2) selection of the relevant set of Principle Components via the Marchenko-Pastur theorem does not outperform arbitrary thresholds of variance explained (95-99%) when the hyperparameters of the machine learning algorithms are finely tuned.