• 中文 (中国)
  • English
  • Hongjian Wang – Urban Computing with Mobility Data: A Unified Approach

    1114wanghj

     

    摘要ABSTRACT:

    Urban computing has gained increasing popularity as an active research topic. It is mostly driven by the fact that we are able to collect various types of urban data in large quantities, including taxi pickups/drop-offs, Tweets from users, environmental sensor measures, POIs, and many more. It is crucial to use these data to understand our city, solve urban issues, and make the city smarter.

    The goal of my research is to develop a unified framework to capture the correlations of heterogeneous data in the urban context. Starting from a preliminary study on estimating the Chicago community level crime with POI and taxi flow. The main takeaway is that the taxi flow defines a new type of region similarity and connects non-adjacent spatial regions. Next, we study the the spatial non-stationary property within these urban features. Namely, a global model does not model the feature relations well. Finally, we address the heterogeneity of the flow-based similarity measures. There are all kinds of pairwise features which define region similarities. We propose an unified model to learn region similarities.

    简介BIO:

    Hongjian Wang is a 5th-year PhD student in College of Information Sciences and Technology at Pennsylvania State University. His adviser is Dr. Zhenhui Li. He interned twice at Twitter Inc in Email Recommendation team (2016) and Ads Prediction team (2017). Before joining Penn State, he received his Bachelor degree from Department of Computer Sciences and Engineering from Shanghai Jiao Tong University in 2013.

    Hongjian’s research interests lie in data mining with a special focus on spatial-temporal data. His research goal is to fuse the heterogeneous urban data to understand urban properties and make our cities smarter. In the past, he worked on several research topics, such as location-based social network relationship mining, travel time prediction, regional crime rate prediction, and region representation learning.

    He has published papers as first author on a number of data mining conferences, including KDD, ICDM, CIKM, SIGSPAITAL, and INFOCOM. Also, he served as reviewer or program committee for a number of data mining venues.