This course explores the intersection of machine learning (ML) and finance, with applications in asset pricing, portfolio optimization, risk management, market microstructure, and more. You’ll learn modern ML techniques on real financial data and build end-to-end pipelines through hands-on projects.
- Format: Lectures • Case studies • Hands-on projects
- Prereqs:
- Background in finance or economics
- ML & applied statistics
- Python programming
- Lec 1 — Intro to Financial ML
Overview; prices-as-predictions; info sets; functional forms; practical challenges. - Lec 2 — Empirical Asset Pricing I
Data; experimental design; linear & penalized models; dimension reduction. (HW1 released) - Lec 3 — Empirical Asset Pricing II
Trees & ensembles; neural nets; alternative data (image/text). - Lec 4 — High-Frequency Prediction
LOB, sampling, order flow, universality, cross impact, spatio-temporal modeling. (HW2 released) - Lec 5 — Optimal Portfolios
Plug-in vs integrated estimation; max-Sharpe regression; trading costs. - Lec 6 — RL in Finance
RL basics; optimal execution; market making; paper discussion; project proposals. - Lec 7 — Generative AI for Finance
GANs/VAEs/diffusion; synthetic backtesting; pitfalls; TailGAN/MARS overview. - Lec 8 — Risk via Graphs
Graph ML; VaR; credit risk (fraud/anomaly/defaults/scoring); systemic risk. - Lec 9 — Large Language Models (LLMs)
Sentiment; report generation; chatbots; reasoning; challenges. - Lec 10 — Interpretability & Ethics
🔍 LIME/SHAP; clustering, t-SNE, graph DBs; bias/fairness; compliance. - Lec 11 — Advanced Topics
Transfer & federated learning; microstructure & stat-arb; backtesting pipeline. - Lec 12 — Case Studies & Labs
HF prediction, volatility forecasting, LLM apps. - Lec 13 — Final Presentations
- HW1 — Cross-Sectional Returns
Linear regression: strengths/limits; predict U.S. equity returns. - HW2 — High-Frequency Modeling
Work with LOB data; replicate & analyze DeepLOB.
Pick one—or propose your own (LLM-centric ideas welcome):
- 📊 Option 1: Generate OHLC Charts
Data: Historical prices (provided)
Outcome: A generative model for OHLC sequences - 📉 Option 2: Stat-Arb with LOB Data
Data: Raw LOBSTER (provided to enrolled students)
Outcome: A data-driven stat-arb strategy - 💡 Propose Your Own (LLMs in Finance):
news summarization • report generation • sentiment for prediction • other LLM apps
Deliverables: proposal, code repo, reproducible notebooks, report, and in-class presentation.
- 🗣️ Participation — 10%
- 📝 Course Assignment I — 15%
- 🧾 Course Assignment II — 25%
- 🎤 Final Project + Presentation — 50%
My sincere thanks to Bryan Kelly, Renyuan Xu, and the many researchers who generously shared slides and teaching materials that informed parts of this course. Their scholarship and openness greatly improved the clarity and rigor of several lectures.
- Financial Machine Learning — Kelly & Xiu (2023)
- Advances in Financial Machine Learning — López de Prado (2018)
- Recent Advances in RL in Finance — Hambly, Xu & Yang (2023)
- The Elements of Financial Econometrics — Fan & Yao (2017)
- The Elements of Statistical Learning — Hastie, Tibshirani & Friedman (2017)