Scholarlink

<p>In E-commerce, price adjustments among substitutable stock-keeping units (SKUs) can trigger cannibalisation and instability. Static models and single-agent reinforcement learning typically fail to capture cross-SKU interactions and fairness considerations. This research proposes a Multi-Agent Reinforcement Learning (MARL) framework using Proximal Policy Optimisation (PPO) for coordinated dynamic pricing. Each SKU operates as an agent, integrating LightGBM demand prediction with Weber-Fechner law behavioral modeling. On simulated electronics e-commerce data, the single-agent baselines optimized for extreme short-term profit at the cost of inventory depletion, whereas the MARL approach achieved sustainable profit and greater system stability. Results demonstrate that coordinated MARL pricing enables more resilient, equitable decision-making under dynamic market conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cognitive multi-agent pricing: an enhanced deep reinforcement learning framework for dynamic SKU competition in E-commerce

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Thanh Ho,
  • Hien Dang,
  • Mai Bui,
  • Nghi Nguyen,
  • Thu Dinh,
  • Hieu Thai

摘要

In E-commerce, price adjustments among substitutable stock-keeping units (SKUs) can trigger cannibalisation and instability. Static models and single-agent reinforcement learning typically fail to capture cross-SKU interactions and fairness considerations. This research proposes a Multi-Agent Reinforcement Learning (MARL) framework using Proximal Policy Optimisation (PPO) for coordinated dynamic pricing. Each SKU operates as an agent, integrating LightGBM demand prediction with Weber-Fechner law behavioral modeling. On simulated electronics e-commerce data, the single-agent baselines optimized for extreme short-term profit at the cost of inventory depletion, whereas the MARL approach achieved sustainable profit and greater system stability. Results demonstrate that coordinated MARL pricing enables more resilient, equitable decision-making under dynamic market conditions.