Workshop Program
All times shown are in UTC-4, which is the eastern daylight timezone (EDT). The workshop starts at 08:00 UTC-4, i.e., 05:00 PDT (Los Angeles), 08:00 EDT (New York), 13:00 CET (Brussels), 20:00 CST (Beijing).
Join Zoom Meeting. Use the following information if needed: Meeting ID: 927 2641 2457 Passcode: 21spatial
| 08:00 | 08:05 | Opening Remarks |
| Session 1: Transportation and Trajectories | ||
| 08:05 | 08:45 | Keynote 1 - Thrill ride: Recommendations in Transportation Systems |
| Eli Safra, Moovit, Israel | ||
| 08:45 | 09:05 | HYPO: Skew-Resilient Partitioning for Trajectory Datasets |
| Giannis Evagorou, Imperial College London, United Kingdom | ||
| Abhirup Ghosh, University of Cambridge, United Kingdom | ||
| Thomas Heinis, Imperial College London, United Kingdom | ||
| 09:05 | 09:15 | Coffee Break |
| Session 2: Areas of Interest | ||
| 09:15 | 09:35 | Which Portland Is It? A Machine Learning Approach |
| Nicole R. Schneider, University of Maryland, College Park, MD, USA | ||
| Hanan Samet, University of Maryland, College Park, MD, USA | ||
| 09:35 | 09:50 | Representation Learning of Urban Regions via Mobility-signature-based Zone Embedding: A Case Study of Seoul, South Korea |
| Namwoo Kim, Korea Advanced Institute of Science and Technology, South Korea | ||
| Yoonjin Yoon, Korea Advanced Institute of Science and Technology | ||
| 09:50 | 10:05 | Finding "Retro" Places in Japan: Crowd-sourced Urban Ambience Estimation |
| Shu Anzai, Nara Institute of Science and Technology, Japan | ||
| Taichi Murayama, Nara Institute of Science and Technology, Japan | ||
| Shuntaro Yada, Nara Institute of Science and Technology, Japan | ||
| Shoko Wakamiya, Nara Institute of Science and Technology, Japan | ||
| Eiji Aramaki, Nara Institute of Science and Technology, Japan | ||
| 10:05 | 10:20 | Visualizing Accessibility with Choropleth Maps |
| David Li, University of Maryland, College Park, MD, USA | ||
| Hanan Samet, University of Maryland, College Park, MD, USA | ||
| Amitabh Varshney, University of Maryland, College Park, MD, USA | ||
| 10:20 | 10:30 | Coffee Break |
| Session 3: Points of Interest 1 | ||
| 10:30 | 11:10 | Keynote 2 - Learning Representations for Hotel Ranking |
| Ioannis Partalas, Expedia, Switzerland | ||
| 11:10 | 11:30 | Spatially and Semantically Diverse Points Extraction using Hierarchical Clustering |
| Naomi Simumba, IBM Research, Japan | ||
| Satoshi Masuda, IBM Research, Japan | ||
| Michiaki Tatsubori, IBM Research, Japan | ||
| 11:30 | 11:40 | Coffee Break |
| Session 4: Points of Interest 2 | ||
| 11:40 | 12:00 | Mining Points of Interest via Address Embeddings: An Unsupervised Approach |
| Abhinav Ganesan, Bundl Technologies (Swiggy), India | ||
| Anubhav Gupta, University of Maryland, College Park, MD, USA | ||
| Jose Mathew, Bundl Technologies (Swiggy), India | ||
| 12:00 | 12:20 | Spatiotemporal Prediction of Foot Traffic |
| Samiul Islam, George Mason University, VA, USA | ||
| Dhruv Gandhi, George Mason University, VA, USA | ||
| Justin Elarde, George Mason University, VA, USA | ||
| Taylor Anderson, George Mason University, VA, USA | ||
| Amira Roess, George Mason University, VA, USA | ||
| Timothy Leslie, George Mason University, VA, USA | ||
| Hamdi Kavak, George Mason University, VA, USA | ||
| Andreas Züfle, George Mason University, VA, USA | ||
| 12:20 | 12:35 | Predicting Customer Poachability from Locomotion Intelligence |
| Syagnik Banerjee, University of Michigan Flint, MI, USA | ||
| Christopher Krebs, University of Michigan Flint, MI, USA | ||
| Neslihan Bisgin, University of Michigan Flint, MI, USA | ||
| Halil Bisgin, University of Michigan Flint, MI, USA | ||
| Murali Mani, University of Michigan Flint, MI, USA | ||
| 12:35 | 12:40 | Closing Remarks |
Keynotes
 |
Keynote 1 - Eli Safra, Moovit, Israel Thrill ride: Recommendations in Transportation Systems Mobility Recommender Systems (MRS) serve passengers by providing recommendations and guidance to travelers and users of both public and private transportation systems. In recent years, the spread and popularity of MRS have grown rapidly due to their ability to plan complex trips based on real-time traffic information. A key aspect of MRS is their ability to effectively compute Expected Time of Arrival (ETA) for vehicles of different types. The ETA can be used for recommending the means of transportation or the driver, in ride-hailing and ridesharing services. However, computing the ETA could be challenging. Moovit is one of the leading MRS and a Mobility as a Service (MaaS) solutions provider. When a company like Moovit starts operating in a new area, it faces the challenge of computing accurate ETA prior to collecting sufficient travel data. Even when external data sources are available, it is not always easy to infer ETA from the data in these sources. In this talk, I will survey the area of MRS and discuss some of the challenges that companies like Moovit face when building MRS, including data collection, storage, update, and usage. I will also present the role of machine learning in MRS and MaaS, e.g., for ETA inference, and how to build mobility recommender systems at scale. BioEli Safra is a Senior Data Scientist at Moovit, with expertise on analyzing geospatial and mobility data. He is also a lecturer of Data Science courses at the Technion – Israel Institute of Technology. Prior to joining Moovit, Eli worked as a Geospatial Expert and Data Scientist at different companies, including Bringg, General Electric, Israel Electricity Company and ESRI. Eli received his Ph.D. from the Technion – Israel Institute of Technology and his M.Sc. from the Hebrew University of Jerusalem. |
||
 |
Keynote 2 - Ioannis Partalas, Expedia, Switzerland
Learning Representations for Hotel Ranking In this talk I will present work on learning item representations from user click session in the hospitality domain and more specifically from the Expedia Group online platforms. I will present in details the proposed neural architecture that leverages side information of items, like attributes and geographic information, in order to learn a joint embedding. I will also explain how it addresses the cold-start problem which is typical in recommendation systems. Results in a downstream task shows that including such structured information improves predictive performance. Finally, I will show through the results of on-line controlled tests that the model generates high quality representations that boost the performance of a hotel recommendation system on Expedia travel platform. BioIoannis Partalas works as Principal Machine Learning Scientist at Expedia Group. His current focus is learning representation in the context of recommendation and ranking systems. Previously he was working as a Research Scientist in Viseo Group, France, on Natural Language Processing. Before that he was an associate researcher in Grenoble-Alpes University working on large-scale classification systems. |
