Coursework from Northwestern’s MS Data Science program
MSDS 451 Financial Machine Learning
This course covered methods to improve statistical robustness in low signal to noise environments, explainable machine learning, and methods to avoid common pitfalls while using algorithms to develop and execute trading strategies. Primary material authored by Dr. Marcos Lopez De Prado and course content created by Dr. Ernest (Ernie) Chan.
Check out the code here
- Fractional differentiation – achieving stationarity without removing all memory from dataset
- Robust feature engineering using hierarchical MDA and MDI (pseudo-randomness does not allow convergence to true feature importances)
- Explainable ML – SHAP, LIME, etc.
- Ergodic Theory
- Combinatorial cross validation (cross validation respecting non-ergodic properties)
- Backtesting strategies, pitfalls, and how to avoid them