Artificial Intelligence-Driven Project Portfolio Optimization Under Deep Uncertainty Using Adaptive Reinforcement Learning

This study proposes an adaptive reinforcement learning (ARL) framework for optimizing project portfolios under deep uncertainty. Unlike traditional static approaches, our method treats portfolio management as a dynamic learning problem. It integrates both explicit and tacit knowledge flows. The framework employs ensemble Q-learning with meta-learning capabilities and adaptive exploration–exploitation mechanisms. We validated our approach across 84 organizations in five industries. The results show significant improvements: 68% in resource allocation efficiency and 52% in strategic alignment (both p < 0.01). The ARL algorithm continuously adapts to emerging patterns while maintaining strategic coherence. Key contributions include (1) reconceptualizing portfolio optimization as learning rather than allocation, (2) integrating tacit knowledge through fuzzy linguistic variables, and (3) providing calibrated implementation protocols for diverse organizational contexts. This approach addresses fundamental limitations of existing methods in handling deep uncertainty, non-stationarity, and knowledge integration challenges.