Predictive modeling in high-frequency trading using machine learning

High-frequency trading (HFT) has transformed financial markets by enabling rapid trade execution and exploiting minute market inefficiencies. This study explores the application of machine learning (ML) techniques to predictive modeling in HFT. Four ensemble boosting methodsAdaptive Boosting, Logic Boosting, Robust Boosting, and Random Under-Sampling (RUS) Boostingwere evaluated using order book data from Euronext Paris. The models were trained and validated on data from a single trading day, with performance assessed using precision, recall, ROC curves, and feature importance analysis. Results indicate that Robust Boosting achieves the highest precision (90%), while Adaptive Boosting and RUS Boosting demonstrate higher recall (94% and 93%, respectively). This research highlights the potential of ML in enhancing HFT strategies, with implications for future trading system developments.