Thea Pan Email and Phone Number

Rates Automated Trading

• Constructed an XGBoost model with 14 time-series and cross-sectional features for stock price prediction; applied a moving window method for model training and prediction; achieved a strategy with a Sharpe Ratio of 2.91• Developed a Python program to evaluate feature effectiveness and identify patterns of feature prediction

• Mined and designed 20+ valid momentum features in C++ after analyzing published research papers; calculated features’ ICs and IRs, and conducted autocorrelation and cross-correlation analysis of the features• Back-tested the strategy after incorporating the selected features, achieving an 18.2% increase in 3-year return

• Conducted parallel comparisons of life model results from Excel and Prophet; detected calculation differences and modified the Prophet coding and Excel formulae, increasing model consistency by 10%