Projects

Selected coursework and modeling projects focused on portfolio construction, return prediction, and out-of-sample evaluation.

30-US-Stock Portfolio Optimization and Backtesting

2026

Financial Market and Instruments

  • Built a Python-based portfolio workflow using 10 years of adjusted daily prices for 30 US blue-chip stocks.
  • Estimated log returns, historical mean returns, and the covariance matrix using in-sample data only.
  • Backtested four strategies: 1/N equal weight, max Sharpe with short selling, max Sharpe without short selling, and minimum variance with short selling.
  • Evaluated equity curve, drawdown curve, annualized Sharpe ratio, and annualized CAR across in-sample and out-of-sample periods.
PythonPortfolio OptimizationBacktestingSharpe Ratio

China A-Share ML Return Prediction and Strategy Backtest

2026

Data Analytics and Machine Learning for FinTech

  • Built a monthly cross-sectional return-prediction workflow for China A-shares using panel data from 2010.01 to 2024.12.
  • Defined 11 predictive factors and completed EDA on missing values, yearly observation counts, factor correlations, and predictive power.
  • Trained and compared Logistic Regression, Lasso, SVM, Random Forest, XGBoost, CNN, RNN, and Transformer models.
  • Constructed a long-only monthly rebalancing strategy based on top-decile predicted probabilities and compared it with the CSI 500 benchmark.
  • Extended validation from a fixed-window split to walk-forward validation to test out-of-sample robustness.
Machine LearningXGBoostTransformerA-ShareBacktesting