Identification of High-Frequency Trading: A Machine Learning Approach

This study aims to develop a probabilistic model using machine learning techniques to identify high-frequency trading (HFT) based on order book data. The model enables precise intraday identifications, addressing the lack of a widely accepted framework for HFT identification and the inconsistencies arising from proxy indicators. Leveraging academic data, the model offers improved consistency and reproducibility for future HFT research. By incorporating fuzzy logic, the probabilistic model allows policymakers greater flexibility in shaping policies. The study utilises data from the BEDOFIH database of the French capital market and develops a robust classification model capable of accurately distinguishing HFT. Additionally, reverse engineering enhances the model’s interpretability by transforming it into an interpretable regression tree without compromising its predictability. This research contributes to advancing HFT research, providing valuable insights, and offering a transferable methodology for identifying HFT in diverse market contexts.