Multi-Armed Bandit Technologies for Automated Inventory Management and Pricing Decisions in Food Retail (Phase I)

用于食品零售中自动库存管理和定价决策的多臂老虎机技术（第一阶段）

• 批准号: 576543-2022
• 负责人: Varro, DanielD
• 金额: $ 3.78万
• 依托单位: McGill University
• 依托单位国家: 加拿大
• 项目类别: Idea to Innovation
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=749758
• 关键词: Multi; Armed; Bandit; Technologies; Automated

Despite food retail generating $14TN+ CAD in revenue globally (2021) and helping feed billions across the world, the market segment largely remains reliant on archaic processes. Today, over one-third of all food is thrown away and most food retailers experience low profit margins in the range of 1-3%. Furthermore, supply chain disruptions (delays) often exacerbate financial and operational challenges, especially, for smaller businesses. Amidst rising food prices, improving food retail business efficiency is an economic necessity. Moreover, poor inventory management leads to unsustainable practices such as excess refrigeration, delivery, and food waste, which alone represents 6-8% of global greenhouse gas emissions. Manual optimization is exceptionally difficult due to large complexity, greatly complicated by variable delays between decisions and their outcome. Our goal is to exploit AI technology for inventory and pricing optimization to improve operational sustainability and efficiency.Recently, we developed novel multi-armed bandit methods and a software package to determine optimal decision-making strategies for situations where delay is observed. Billions of simulations validated that our methods can offer over 40% better decision-making performance than traditional multi-armed bandit methods in the presence of delay. The methods work by evaluating possible decision outcomes and using them to calculate expected outcomes.The NSERC I2I grant will allow us to augment and validate our technology in a real market scenario in food retail to better address retail customer needs. Our technology will improve both human-in-the-loop and autonomous decision-making for sustainable inventory management by providing effective recommendations and explanations concerning purchase amounts and pricing (including discounts for consumers). This will lead to improved margins in food retail and lower prices for consumers, as well as more sustainable practices which minimize food waste, cut down on unnecessary deliveries, and reduce energy consumption.

尽管食品零售业在全球收入（2021年）中产生了14tn+ CAD，并帮助全球数十亿美元的饲养，但市场领域仍在很大程度上依赖于古流程。如今，超过三分之一的食物被扔掉了，大多数食品零售商的利润率低于1-3％。此外，供应链中断（延误）通常会加剧财务和运营挑战，尤其是对于小型企业而言。在粮食价格上涨中，提高粮食零售业务效率是经济上的必要性。此外，库存管理差会导致不可持续的做法，例如过量的制冷，交付和食物浪费，仅占全球温室气体排放的6-8％。 由于复杂性很大，手动优化非常困难，因此决策及其结果之间的延误非常复杂。我们的目标是利用AI技术进行库存和定价优化，以提高运营的可持续性和效率。实际上，我们开发了新型的多武器强盗方法和软件包，以确定观察到延迟的情况的最佳决策策略。数十亿个模拟证实了我们的方法可以比传统的多武器匪徒在存在延迟的情况下提供40％以上的决策性能。这些方法通过评估可能的决策成果并使用它们来计算预期的结果来起作用。NSERCI2I赠款将使我们能够在食品零售的真实市场情况下增强和验证我们的技术，以更好地满足零售客户需求。我们的技术将通过提供有关购买金额和定价的有效建议和解释（包括消费者的折扣）来改善可持续库存管理的人类和自主决策。这将导致食品零售的利润率提高，消费者的价格降低，以及更可持续的做法，这些实践可以最大程度地减少食物浪费，减少不必要的交付并减少能源消耗。