← BACK TO PROJECTS
Regime-Aware Portfolio Optimization
#Quantitative Finance
#Hidden Markov Models
#Convex Optimization
#CUSUM Sampling
Executive Overview
Integrated Gaussian Hidden Markov Models (HMMs), Random Forest meta-models and convex portfolio optimization across 503 equities (2018–2024).
Key Methodological Innovations
- CUSUM Sampling & Triple Barrier Method: Sampled informative event timestamps, filtering microstructural noise with 5-day horizon and 1.5x dynamic ATR thresholds.
- Purged 5-Fold Cross-Validation: Implemented with 1% embargo periods preventing information leakage.
- Regime-Conditional Allocation: Transitioned dynamically between Maximum Sharpe (bullish) and Hierarchical Risk Parity (high-volatility).