讲座题目：Improve Supply Chain and Logistics Forecasting by Machine Learning Techniques

时间：2018年7月4日（周三）上午8:30-10:00

地点💯：主教615会议室

演讲人：Hing Kai Chan教授

Professor Chan has published over 100 peer-reviewed academic articles and (co-)edited several special issues for reputable international journals. His publications appear in Production and Operations Management, European Journal of Operational Research, various IEEE Transactions, Decision SupportSystems, International Journal of Production Economics, International Journal of Production Research, among others. He has been the co-editor of Industrial Management & Data Systems (SCI-indexed)since 2014. He was the Associate Editor of the IEEE Transactions on Industrial Electronics (SCI-indexed) from 2009 to 2015, and the Associate Editor of the IEEE Transactions on Industrial Informatics (SCI-indexed) from 2014-2017. Professor Chan also serves as an Editorial Board Member (or similar) in a number of journals such as Transportation Research Part E: Logistics and Transportation Review (SCI-indexed), Online Information Review (SCI-indexed).

Abstract:

Data driven research has been very popular in the last few years. Machine learning is an important element of this strand of research. In this presentation, machine learning approaches are employed to make prediction on two supply chain and logistics applications: the demand of healthcare products and the container throughput of a port. The two applications share some similarities: (i) A number of socio-economic data are extracted and considered as the inputs to the prediction model; (ii) Some of these parameters are not directly linked to the prediction in previous studies; and (iii) A correlation or cluster analysis is first conducted to reveal the relationship of those parameters. In order to verify the models, real-life data were collected and employed. Results indicate that incorporating such data, some of which are apparently irrelevant to the prediction, could improve the forecasting accuracy. It is also found that various machine learning approaches are also useful in the prediction, compared to some traditional time-series models, such as ARIMA. That being said, machine learning approaches are not necessarily the best method. In this presentation, more details will be discussed.