Machine Learning for Quantitative Trading

This research collection presents two complementary papers on advanced machine learning applications in quantitative trading. The first paper introduces a novel approach for predicting iceberg order execution using XGBoost models, achieving 79% precision by analyzing market microstructure patterns. The second paper systematically explores hyperparameter optimization strategies for these prediction models, revealing that shorter training windows (just two time periods) consistently outperform longer historical datasets across all tested algorithms. Together, these works demonstrate how carefully optimized machine learning models can extract predictive signals from market data, quantify execution uncertainty, and create adaptive trading strategies that respond dynamically to changing market conditions. The included model comparison reports provide detailed optimization results for XGBoost, LightGBM, Random Forest, and Logistic Regression, with the surprising finding that simpler Logistic Regression models achieved the highest overall performance (0.68990) when properly regularized.