Abstract

The growing competition in the retail market pushes retailers to optimize their sales and marketing strategies. In particular, in sectors where the profit margins are more restricted and customer loyalty depends heavily on the prices offered, in fact, understanding consumer reactions to sales promotions and providing them with the right deal at the right time is critical for retailers to survive. In this study, we propose an agent-based model that uses the Belief Desire Intention (BDI) concept to model how customers react to the various promotional offers they receive, and our study involves a seller agent that dynamically learns the appropriate promotion to give these customers on a customer-specific basis. Using empirical research findings in the literature, we developed the BDI part of the model based on the “Big Five-Factor Model”. Apart from this, we used the Q-learning algorithm to train the seller agent. Furthermore, one of our goals in this study is to show that learning-based decision-making agents can be more competitive on the market than rule-based decision-making agents. We therefore added a rule-based agent to the model and compared its efficacy to that of the learning-based decision-making agent. The model was tested in an artificial market through a series of experiments. The experiment results show that the proposed model can be used in actual applications to automate sales promotion decisions.

中文翻译：

---

基于Agent仿真的双向促销建模

摘要

零售市场日益激烈的竞争促使零售商优化其销售和营销策略。特别是在利润率更受限制且客户忠诚度很大程度上取决于所提供的价格的行业中，事实上，了解消费者对促销活动的反应并在正确的时间向他们提供正确的交易对于零售商的生存至关重要。在这项研究中，我们提出了一种基于代理的模型，该模型使用信念欲望意图（BDI）概念来模拟客户对他们收到的各种促销优惠的反应，我们的研究涉及一个卖家代理，它动态地学习适当的促销活动来给予这些优惠客户根据客户具体情况。利用文献中的实证研究结果，我们基于“大五因素模型”开发了模型的 BDI 部分。除此之外，我们还使用 Q-learning 算法来训练卖家代理。此外，我们在这项研究中的目标之一是证明基于学习的决策代理比基于规则的决策代理在市场上更具竞争力。因此，我们在模型中添加了一个基于规则的代理，并将其功效与基于学习的决策代理的功效进行了比较。该模型通过一系列实验在人工市场上进行了测试。实验结果表明，所提出的模型可以在实际应用中用于自动化促销决策。