Bridging behavioral insights and quantitative finance: AI-powered Black–Litterman framework with technical and sentiment signals

Aiming to bridge quantitative finance with behavioral economics, this study harnesses artificial intelligence (AI) to integrate high-quality market sentiment into portfolio optimization. It evaluates the performance of the Black–Litterman (BL) asset allocation model, incorporating investor views generated from state-of-the-art deep learning (DL) models. These models are trained on three distinct datasets—technical (TD) derived from historical US sectoral ETF prices, sentiment (SD) obtained from Refinitiv’s MarketPsych Analytics (LSEG), and their combination (TSD). The proposed framework replaces subjective expert views with data-driven forecasts to enhance accessibility for retail investors. Portfolios are constructed with daily rebalancing based on DL-forecasted prices and account for transaction costs under different market regimes and risk aversion rates. The findings reveal that BL models incorporating the integrated TSD with lower risk aversion (λ=1) significantly outperform those based on TD, SD, or traditional benchmarks, underscoring the robustness of combining technical and sentiment signals for view generation and highlighting its effectiveness for growth-oriented strategies. Under normal market conditions, TD and SD-based portfolios exhibit comparable average performance on risk-adjusted evaluation metrics; however, in high-volatility regimes, TD-based portfolios consistently outperform their SD counterparts on average. This study advocates for TSD-based, DL-enhanced BL models with lower risk aversion as a robust strategy in dynamic market environments, offering practical guidance for retail investors and insights for policymakers on harnessing AI to strengthen financial decision-making.