Computational finance undergraduate focused on transparent, reproducible empirical research in equity factor investing and credit-risk machine learning.
professional email: dingwenqi6@gmail.com
- Empirical factor research
- Quantitative finance
- Credit-risk machine learning
- Python: pandas, NumPy, scikit-learn, statsmodels, LightGBM
- Research: factor IC/ICIR, quantile returns, portfolio backtesting, performance attribution
- Tools: Git, GitHub Actions, Streamlit, JupyterLab
- Data: deterministic simulated panels (default) and an experimental Tushare A-share adapter
My main public research repository is quant-factor-research. The other projects demonstrate complementary interfaces and forecasting methods. The projects most relevant to an application depend on the research question and the receiving program.
| Project | Role in the portfolio | What to review |
|---|---|---|
| Reproducible Multi-Factor Equity Research | Flagship research project | Synthetic pipeline validation plus official Fama-French factor and momentum-decile evidence, dependence-aware inference, failure analysis, and reproducible artifacts |
| Quant Factor Lab | Interactive demo | A clearly labeled synthetic Streamlit interface for exploring the flagship project's factor workflow |
| Financial Time-Series Baseline | Forecasting baseline / scaffold | Synthetic harness plus a retrospective official aggregate-market Ridge comparison; historical availability is not verified and this is not a complete TimeCAP reproduction |
My factor research repository uses a deterministic simulated stock panel to exercise timing checks, transaction costs, and reproducibility. The stock-level simulation does not implement a train/validation/test split, and its results are pipeline evidence, not market alpha. A separate module analyzes official aggregate factor returns; it does not validate the stock-selection strategy. The A-share adapter is experimental and requires further point-in-time and execution checks.
- Keep targets and future information outside feature construction.
- Compare models and strategies with transparent baselines.
- Report weak and negative findings alongside favorable results.
- Make assumptions, data provenance, costs, and claim boundaries inspectable.
- Treat synthetic experiments as pipeline validation, not market evidence.
- Point-in-time real data integration for China A-shares and US equities
- Walk-forward out-of-sample evaluation and factor-decay monitoring
- Multiple-testing controls and deflated performance measures
- Credit-risk model interpretation and validation
Python is my primary research language. The repositories above include setup instructions, automated tests, and generated outputs so reviewers can reproduce the work rather than relying on screenshots or unsupported performance claims.