讲座题目：When to and When Not To Make a Recommendation? A Product-Centric Approach to Optimize the Timing Decision of Online Recommendations

报告人：常象宇 贝斯特bst3344游戏副教授

报告时间：2019年12月23日星期一下午15：00-16：30

报告地点：管院311会议室

报告内容: Recommending the right product at the right time for consumers is the goal of modern online recommendation systems. However, consumers who shop in e-commerce often complain about a phenomenon that the product which you just bought is recommended again. There are two possibilities for this kind of complaint. First, the recommended items do not need to be purchased repeatedly. Second, the recommended item is repurchable, but the recommendation time is not proper. To optimize current recommendation systems and improve consumer’s satisfactions, this paper studies the following problems:

1. Why recommend products that you just purchased?

2. What type of products are prone to repurchase? What type of products are not prone to repurchase?

3. What is an optimal time interval for the recommendation?

4. Can we optimize the current recommendation system so that we can reduce customer complaints while still keep the purchase rate?

To this end, we introduce the consumer-based analysis which has been well studied in the marketing science into the machine-learning-based recommendation systems to handle the above problems.

报告人简介：常象宇， 贝斯特bst3344游戏副教授，华盛顿大学西雅图分校工业与系统工程系客座副教授。2017年入选陕西省高等学校优秀青年学者支持计划。研究主要集中在统计机器学习，及其在管理问题上的应用。曾在统计学期刊AOS，JOE，SS，EJS等；机器学习期刊JMLR，TNNLS，TC，TSP等发表论文三十余篇。同时也曾是人工智能与数据挖掘相关会议：ICML，AAAI，IJCAI，SDM，ICDM，ICDS 等的程序委员会委员。