Diploma and Master Theses (authored and supervised):

C. Zuo:
"Extraction and Visual Analysis of Negative Traffic Events from Weibo Data";
Supervisor: L. Ding, G. Gartner; Department of Civil, Geo and Environmental Engineering, TU München, 2017; final examination: 2017-09-25.

English abstract:
Social media has become prevalent in the last decade. People publish various topics, for instance food, work, traveling, sports, health, on a variety of social media platforms on a daily basis, which reflect a wide range of socioeconomic phenomena and human behaviors. Social
media data has been widely studied in many research fields and a variety of methods haven been proposed to extract meaningful information that can be applied in domains like crisis management, sentimental analysis, etc.
This thesis aims to propose event extraction and visualization methods with a focus on traffic-related topics, especially negative traffic topics, from social media data. To fulfill it, we propose a methodology for the traffic event extraction and visualization. Firstly, based on the identification and classification of representative negative traffic topics, the author proposed an iterative text mining method to identify negative traffic events information from social media data. Secondly, from the extracted traffic topics, the author performed a set of clustering and aggregation methods to derive high-level information for further hotspot analysis. Thirdly, the author proposed several visual analytic techniques to explore the negative spatiotemporal traffic patterns of
both individual traffic data and the aggregated data. A web-based interactive visual analytic system integrating animated time graph, scatter plot maps, heatmaps, chord diagrams and alluvial diagrams are developed to visualize the significant hotspots and relationships between negative traffic event topics.
We apply the methodology to one-year Weibo data in Shanghai in 2014 to investigate the negative traffic events patterns. The experiment shows significant spatiotemporal negative traffic patterns, for instance hourly/daily patterns, divers spatial distributions at city center and satellite cities, and also possible correlations among different types of events.

Social Media, Event Extraction, Visual Analytics, Spatiotemporal Mining, Text Mining

Created from the Publication Database of the Vienna University of Technology.