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.

尽管食品零售在全球(2021年)创造了超过14万亿加元的收入，并帮助养活了世界各地的数十亿人，但该细分市场在很大程度上仍然依赖于陈旧的流程。今天，超过三分之一的食品被扔掉，大多数食品零售商的利润率都很低，在1%到3%之间。此外，供应链中断(延误)往往加剧财务和业务挑战，特别是对较小的企业而言。在食品价格不断上涨的情况下，提高食品零售企业的经营效率是经济上的必然。此外，糟糕的库存管理导致了不可持续的做法，如过度冷藏、送货和食物浪费，仅这些就占全球温室气体排放量的6%-8%。人工优化非常困难，因为它非常复杂，决策和结果之间的可变延迟使其变得非常复杂。我们的目标是利用人工智能技术进行库存和定价优化，以提高运营的可持续性和效率。最近，我们开发了新的多臂强盗方法和一个软件包，以确定在观察到延迟的情况下的最优决策策略。数十亿次的仿真结果表明，在存在时延的情况下，我们的方法比传统的多臂强盗方法可以提供超过40%的决策性能。这些方法通过评估可能的决策结果并使用它们来计算预期结果。NSERC I2I授权将允许我们在食品零售的真实市场场景中增强和验证我们的技术，以更好地满足零售客户的需求。我们的技术将通过提供有关采购金额和定价(包括消费者折扣)的有效建议和解释，改善可持续库存管理的人在环和自主决策。这将提高食品零售的利润率，降低消费者的价格，以及更可持续的做法，将食物浪费降至最低，减少不必要的送货，并减少能源消耗。