华南理工大学工商管理学院“海外优青分论坛”，旨在面向全球邀请拥有不同学术背景的青年才俊，围绕管理学科相关领域国际科学前沿、热点研究领域以及行业产业的技术问题等展开探讨和交流。通过这个平台，互相启迪、开拓视野，增强国际交流与合作，促进双方共同发展。

一、论坛时间

2023年12月22日下午3点

二、地点

华南理工大学五山校区12号楼105会议室

三、论坛议程

欢迎广大师生参加！

华南理工大学工商管理学院

2023年12月14日

附：

报告1：数据驱动的智能维护管理（刘斌博士英国思克莱德大学）

内容摘要：

This talk presents dada-driven decision-making for intelligent maintenance management, focusing on failure prediction, and maintenance optimisation. Firstly, I will discuss extending the applicability of deep learning algorithms for failure prediction with limited time series data. Data augmentation methods are leveraged to increase the sample size to accommodate the deep learning algorithms. Different from existing approaches that indiscriminately blend synthetic and real data without considering the selection or weighting of synthetic samples, our novel method involves generating, selecting, and reweighting synthetic samples to enhance prediction accuracy. A case study of failure prediction in a wastewater treatment plant will be used to illustrate the effectiveness of the proposed method.

Subsequently, I will discuss data-driven maintenance management, including end-to-end preventive maintenance, and the development of online reinforcement learning for maintenance decision-making. The end-to-end approach utilizes historical data to directly suggest the optimal preventive replacement time without assuming the underlying failure distribution. In addition, Markov decision processes and online reinforcement learning will be discussed for condition-based maintenance modelling and optimization.

报告人简介：

刘斌，博士，英国思克莱德大学管理科学系高级讲师（副教授）。他在浙江大学获得了自动化工程学士学位，并在香港城市大学获得了系统工程博士学位。刘斌博士的研究领域涵盖风险和可靠性分析、智能维护管理以及数据驱动的决策分析。他的论文发表在多个跨学科领域期刊上，如European Journal of Operational Research，Automatica，IISE Transactions和IEEE Transactions等期刊。

报告2：Deep Neural Newsvendor（韩锦晖博士 加拿大多伦多大学）

内容摘要：

We consider a data-driven newsvendor problem, where one has access to past demand data and the associatedfeature information. We solve the problem by estimating the target quantile function using a deep neuralnetwork (DNN). The remarkable representational power of DNN allows our framework to incorporate orapproximate various extant data-driven models. We provide theoretical guarantees in terms of excess riskbounds for the DNN solution characterized by the network structure and sample size in a non-asymptoticmanner, which justify the applicability of DNNs in the relevant contexts. Specifically, the convergence rateof the excess risk bound with respect to the sample size increases in the smoothness of the target quantilefunction but decreases in the dimension of feature variables. This rate can be further accelerated whenthe target function possesses a composite structure. Compared to other typical models, the nonparametricDNN method can effectively avoid or significantly reduce the model misspecification error. In particular,our theoretical framework can be extended to accommodate the data-dependent scenarios, where the data-generating process is time-dependent but not necessarily identical over time. Finally, we apply the DNNmethod to a real-world dataset obtained from a food supermarket. Our numerical experiments demonstratethat (1) the DNN method consistently outperforms other alternatives across a wide range of cost parameters,and (2) it also exhibits good performance when the sample size is either very large or relatively limited.

报告人简介：

韩锦晖，博士，多伦多大学罗特曼管理学院博士后，师从胡明教授。2021年获香港中文大学统计学博士学位，并在2021-2022年在香港中文大学开展博士后研究。主要研究方向为数据驱动和随机模型在运营管理中的应用、运筹与经济金融学交叉研究、运营风险管理及金融工程，研究成果发表在Production and Operations Management，European Journal of Operational Research，SIAM Journal on Control and Optimization等期刊。