Business optimisation is the process of finding and implementing efficient and cost-effective means of operation to bring a competitive advantage for businesses. Synthesizing problem formulations is an integral part of business optimisation which is centred around human expertise, thus with a high potential of becoming a bottleneck. With the recent advancements in Large Language Models (LLMs), human expertise needed in problem formulation can potentially be minimized using Artificial Intelligence (AI). However, developing a LLM for problem formulation is challenging, due to training data requirements, token limitations, and the lack of appropriate performance metrics in LLMs. To minimize the requirement of large training data, considerable attention has recently been directed towards fine-tuning pre-trained LLMs for downstream tasks, rather than training a LLM from scratch for a specific task. In this paper, we adopt this approach and propose an AI-Copilot for business optimisation by fine-tuning a pre-trained LLM for problem formulation. To address token limitations, we introduce modularization and prompt engineering techniques to synthesize complex problem formulations as modules that fit into the token limits of LLMs. In addition, we design performance evaluation metrics that are more suitable for assessing the accuracy and quality of problem formulations compared to existing evaluation metrics. Experiment results demonstrate that our AI-Copilot can synthesize complex and large problem formulations for a typical business optimisation problem in production scheduling.

翻译：商业优化是发现并实施高效且具有成本效益的运营方式，从而为企业带来竞争优势的过程。问题建模的综合是商业优化中依赖人类专业知识的核心环节，因此极易成为瓶颈。随着大语言模型（LLMs）的最新进展，利用人工智能（AI）有望最小化问题建模所需的人类专业知识。然而，由于训练数据需求、Token限制以及LLMs中缺乏合适的性能评估指标，开发用于问题建模的LLM具有挑战性。为减少对大规模训练数据的需求，近期研究重点转向微调预训练LLM以完成下游任务，而非从零开始训练特定任务的LLM。本文采用这一方法，通过微调预训练LLM进行问题建模，提出面向商业优化的AI辅助系统（AI-Copilot）。针对Token限制问题，我们引入模块化设计与提示工程（prompt engineering）技术，将复杂问题建模分解为符合LLM Token限制的模块。此外，我们设计了比现有评估指标更适用于衡量问题建模准确性与质量的性能评估指标。实验结果表明，该AI-Copilot能够为生产调度中典型的商业优化问题综合出复杂且大规模的问题建模方案。