Agent-driven big data interaction empowers personalised marketing efficiency

This paper proposes a cognition-aware hierarchical reinforcement learning framework to solve the decision-making inefficiency problem caused by ignoring the user's cognitive status in traditional personalised marketing systems.By formalising cognitive load theory as a reward function and constructing an agent that collaborates metacognition and executive layer, we achieve dynamic optimisation of user cognitive experience while accurately recommending.Experiments based on the real user behaviour dataset of Alibaba Mobile show that the proposed model is significantly better than the state-of-the-art baselines such as deep interest evolution network and deep reinforcement learning for news recommendation in key indicators such as normalised discounted cumulative gain at 10 and conversion rate (p < 0.01).Ablation experiments further verify the necessity of cognitive components and hierarchical architecture.