In this project, supervised by SEB, analysts Ludvig Cederlund, Sofie Melander and Malcolm Cederbaum investigate how the EUR/SEK exchange rate can be predicted using market data along with proprietary trade data from SEB. Utilizing the XGBoost model, they work on a 10-minute frequency to perform binary classification of short-term price direction.
The analysis finds that a substantial share of short-term price movements can be explained trough the SEB customer flow data, capturing unique information about localized market sentiment, order imbalances, and liquidity demand not fully reflected in public market variables. This underscores the value of internal flow data as a competitive advantage in quantitative FX prediction.