A self-adjusting Decision Support System for strategic project portfolio prioritization using integrated AHP and reinforcement learning

This study proposes a novel Adaptive Decision Support System (ADSS) that integrates Analytic Hierarchy Process (AHP) and Q-Learning. The system is designed to optimize project portfolio management under dynamic and uncertain conditions. The framework addresses a critical challenge in strategic decision-making by combining static managerial preferences captured through AHP with a reinforcement learning agent that adaptively adjusts its actions in response to real-time environmental feedback. AHP-derived weights for Key Performance Indicators (KPIs) are incorporated into a strategically designed reward function that guides the RL agent, operating under an ε-greedy policy, to learn an optimal long-term allocation strategy. This ensures that adaptive decisions remain aligned with organizational priorities while responding flexibly to operational changes. The framework was validated via dynamic NetLogo simulations, demonstrating a 21% improvement in cumulative portfolio value and strategic alignment compared to conventional non-adaptive approaches. Results indicate that the ADSS consistently generates high-quality, data-driven decisions, enhancing resilience, resource allocation, and long-term portfolio performance. This framework provides actionable insights for managers and decision-makers seeking to integrate strategic goal-setting with adaptive operational decision-making in complex and uncertain environments.