My winning solution to the CFM 2020 data science challenge.
Where will the next trade take place?
Given past trades and order books snapshots from a set of six trading venues, the goal of this data science challenge was to predict the venue of the next trade.
My solution relied on gradient boosting trees with handcrafted features to account for liquidity fragmentation, order flow autocorrelation, intraday patterns and stock specificities.
Presentation of the challenge by organizers available at this address: https://www.college-de-france.fr/site/stephane-mallat/Challenges-2020.htm
- add slides
- add code