Integrating minds: An ensemble approach to portfolio optimization

Estimation errors in covariance matrices significantly undermine the effectiveness of portfolio optimization. To address this challenge, we introduce a novel ensemble approach that builds upon the strategy of “divide-and-conquer” and the principle of “two minds are better than one”. This simple approach forms a strong ensemble portfolio by constructing and aggregating numerous base models. The proposed method enhances estimation accuracy and stability through dimension reduction by asset grouping, noise mitigation via cross-model averaging, and complementary information from diverse ensemble learners. Comparative evaluations exhibit the superior performance of our method with respect to portfolio risk, trading turnover, and net Sharpe ratios.