I am a second-year PhD student in College of Information Studies at University of Maryland, College Park, advised by Vanessa Frias-Martinez. Prior to joining UMD, I got my master degree from Deparment of Information Management at Peking University under the guidance of Jun Wang and Win-bin Huang, and got my bachelor degree from School of Information Management at Sun Yat-Sen University.
My general research interest is to adopt data science approaches to solve questions of Smart Cities and Communities. While I am trying to narrow down my research topic, urban transportation issues seems to be a good start. I am also interested in learning and practicing machine learning and visualization techniques.
Cycling Safety Maps: A Data-Driven Approach to Cycling Safety
Cycling safety around the city is an important factor for people to decide to ride a bike or not. Also, transportation department invest significant resources into building cycling-friendly environment. But traditional approach to compute safety maps require extensive information which is expensive to collect. Therefore this project aims at building such a map automatically. First we collect ground truth cyclist-perceived safety level with this rating platform. Meanwhile, we collect various data, such as Point-of-Interest from OpenStreetMap and FourSquare, crashing data and moving violation from Open Data DC, based on which segment-level features are extracted. These kinds of data are often easy to obtain. Then we train probabilistic models on these features and ground truth so that we can computed cycling safety map effortlessly. Finally, we provide insights about how cyclist perceive cycling safety in terms of easy-access data.
User Behavior and Profile Mining Based on Cellular Mobile Data Service Logs
Cellular mobile data service logs typically are tuples of user, base station, timestamp and application. These logs can reflect user needs in real time, including social interaction, shopping and dining. Therefore, our goal is to understand user behavior and needs and to obtain comprehensive user profiles on both micro and macro perspectives. In this project, a framework for data-driven mobile user behavior analysis, namely was proposed. Following this framework, we started analyzing and understanding user behavior, and focused on spatial behavior modeling including periodical pattern and moving status based on a 0.5TB dataset. Periodical probability matrix of locations was constructed to identify user’s home and other frequent places. Moving status pattern was exploited to tell moving behavior in leisure time and typical commuting behavior.
Title-based Academic Papers Aggregation System
When you are new to a field, have you ever had a hard time trying to figure out classic concepts and methods in that field? In this project, I built a system to navigate technical academic papers based on automatically extracted concepts in technical academic papers. This system makes use of the linguistic and structure characteristics of titles and outlines in papers to extract concepts and their relationship, which are used to build the content-based navigation and concept network. Here is a demo of the system.
Win-bin Huang, Shanchuan Xu, Jiahui Wu, Jun Wang (2016). Construction of Mobile User Profile. Journal of Modern Information, 36(10), 54-61.(in Chinese)
Win-bin Huang, Jiahui Wu, Shanchuan Xu, Jun Wang(2016). Data-driven Mobile User Behavior Analysis Framework and Technology. Information Science, V34(7):14-20. (in Chinese)