Dynamic Spatio-Temporal Graph Attention Networks with Nested Clustered Optimization for Multi-Asset Portfolios
Abstract
Nested Clustered Optimization (NCO) insulates asset allocation from Markowitz covariance ill-conditioning. However, classical tree clustering relies on static pairwise Pearson correlation, failing to capture time-varying, asymmetric downside risk contagion across multi-asset markets. We propose a unified deep graph machine learning architecture: Dynamic Spatio-Temporal Graph Attention Networks with Nested Clustered Optimization (ST-GAT-NCO) for multi-asset tactical allocation. The framework extracts temporal momentum and downside risk dynamics via causal dilated 1D-convolutions, then deploys a multi-head Graph Attention Network (GAT) over dynamic market topologies to learn latent spatial node representations. These learned embeddings define an angular metric distance space that guides two-stage downside NCO optimization under institutional constraints and turnover smoothing. Evaluated over a 73-month non-anticipative walk-forward out-of-sample window (January 2020 to January 2026) across 13 global asset classes net of 10 bps transaction costs, ST-GAT-NCO achieves a 5.32% CAGR and 37.06% cumulative net return, delivering an excess Sharpe ratio of 0.30 and a Sortino ratio of 0.48. This substantially outperforms classical Minimum Variance (3.68% CAGR, 0.08 Sharpe), Hierarchical Risk Parity (2.65% CAGR), and Plain Symmetric NCO (4.11% CAGR, 0.15 Sharpe), while restricting maximum drawdown to -13.07% (versus -19.44% for Equal-Weight and -16.36% for Downside NCO) under a controlled one-way turnover of 15.31%/month. Regime analysis confirms active capital preservation across the 2020 COVID-19 crash, 2022 monetary tightening, and 2023-2026 disinflation rallies.
Figures & tables
| Strategy | CAGR | Ann. Vol | Down. Vol | Sharpe | Sortino | Max DD | Calmar | Total Ret | Turnover |
|---|---|---|---|---|---|---|---|---|---|
| (%) | (%) | (%) | Ratio | Ratio | (%) | Ratio | (%) | (%/mo) | |
| Equal-Weight ( ) | 11.40 | 11.04 | 6.11 | 0.72 | 1.31 | -19.44 | 0.59 | 92.84 | 0.07 |
| Min-Variance (MVO) | 3.68 | 2.62 | 1.63 | 0.08 | 0.13 | -3.24 | 1.13 | 24.56 | 6.81 |
| Hierarchical Risk Parity (HRP) | 2.65 | 0.73 | 0.52 | -1.14 | -1.59 | -0.53 | 4.96 | 17.21 | 4.61 |
| Plain NCO (Symmetric) | 4.11 | 4.58 | 2.65 | 0.15 | 0.26 | -8.58 | 0.48 | 27.75 | 22.40 |
| Downside NCO (Semi-Cov) | 4.26 | 5.74 | 3.59 | 0.16 | 0.25 | -16.36 | 0.26 | 28.92 | 20.91 |
| Macro Stress Regime / Period | Equities | Fixed Inc. | Real Assets | Cash/Curr. | Crypto | Monthly Turn. | Effective | Max Single |
|---|---|---|---|---|---|---|---|---|
| (%) | (%) | (%) | (%) | (%) | (%/mo) | Assets ( ) | Pos. (%) | |
| Full OOS Period (2020–2026) | 10.64 | 33.48 | 21.38 | 32.48 | 2.01 | 15.31 | 8.24 | 19.82 |
| COVID-19 Liquidity Shock (2020) | 7.74 | 42.71 | 16.75 | 28.56 | 4.24 | 9.84 | 8.68 | 17.19 |
| Post-Pandemic Growth Rally (2021) | 13.19 | 43.92 | 12.93 | 28.08 | 1.87 | 13.17 | 8.69 | 17.80 |
| Global Monetary Tightening (2022) | 12.12 | 32.47 | 23.36 | 32.05 | 0.00 | 16.27 | 8.23 | 19.67 |
| Disinflation & AI Rally (2023–2024) | 4.43 | 26.52 | 25.71 | 41.67 | 1.68 | 17.78 | 6.67 | 23.83 |
Portfolio Optimization
Portfolio optimization selects asset weights to balance expected return, risk, diversification, and trading constraints.
See all 73 papers