In this project, supervised by OQAM, the analysts Olivia Lahtinen, Behdad Nikfarjam, and Douglas Eklund explore whether machine learning techniques can uncover pricing inefficiencies in the Nordic covered bond market. Using a Random Forest model, they aim to predict movements in five-year Swedish covered bond yields.
The model demonstrated encouraging results in the Swedish market, achieving a directional accuracy of 58.84% and an R-squared of 0.1843, indicating meaningful predictive ability. These results suggest that machine learning can support improved investment decision-making in a traditionally stable and under-explored segment of the fixed-income market.