Empirical Asset Pricing via Ensemble Gaussian Process Regression
创建于 更新于
摘要
Authors introduce an ensemble Gaussian Process Regression (GPR) to predict conditional expected stock returns and to quantify epistemic uncertainty, demonstrating superior out-of-sample predictive R-squared and information coefficient versus linear benchmarks, and showing that uncertainty-aware portfolios (UW, PUW) deliver large economic gains (higher Sharpe) compared to EW/VW strategies; the ensemble also reduces GPR computational burden and supports online learning [page::0][page::3][page::19][page::24]
