Virtual Seminar Series: Yicheng Song, Carlson School of Management, University of Minnesota
December 4, 2020
2:00 PM - 3:30 PM
Join us for a virtual seminar this semester with Yicheng Song is an Assistant Professor of Information and Decision Sciences at Carlson School of Management, University of Minnesota.
Please use the link below to join via Zoom.
Meeting ID: 975 3576 7242
Password: IDStalk20
Talk: Short-Lived Item Recommendation
Abstract: In a highly dynamic market (e.g., online news and flash sales), new items continuously flow into the marketplace and fade out within a short period, making the item recommendation a big challenge. The absence of user-item interactions for the newly added items brings in the cold-start problem, while the recommender systems need to promptly update the user and item representations to incorporate the most recent interactions in response to the short lifespan of items and shifting preference of users. To address these challenges, we propose a Dual Recurrent Neural Network (DRNN), which aims to learn effective initial representations of items, and efficiently update representations of users/items as new interactions occur. Empirical experiments on click-stream data from an online flash sale retailer show that compared with the best state of the art benchmark, DRNN performs 1.8 times better in prediction accuracy on cold-started items, 33% better on overall next-item interactions, and 13% better in matching the diversity of items users are seeking to. We further explore the item representations by unveiling the dynamic updating process of item representations in DRNN still follows the principle of collaborative filtering and uncovering latent market structure from item embedding to help managers better understand product/category position and distribution.
Bio: JYicheng Song is an Assistant Professor of Information and Decision Sciences at Carlson School of Management, University of Minnesota. His research focuses on investigating digital user’s decision journey and the corresponding firm strategies using inter-discipline approaches: Machine Learning, Bayesian Modeling, Economic Structural Modeling and Reinforcement Learning. His research has been published in a number of leading academic journals in information systems and computer science, including management science, IEEE Transactions on Multimedia and ACM Transactions on Intelligent Systems and Technology. He holds a PhD in information systems from Boston University and PhD in Computer Science from Chinese Academy of Sciences.
Date posted
Nov 30, 2020
Date updated
Nov 30, 2020