Spatio-temporal data containing multi-variate attributes are of perticular interest for us, for example Weibo (Chinese Twitter) check-in data, Taxi data, as well as data from telco operators. These data sourse is very unique, as they contain large quantities of users and rich spatio-temporal information of peoples' all-in-one behavior in daily lives. It provides unprecedent opportunity to study urban movement patterns. Depending on the specific problems, different mining techniques can be applied, and normally standard algorithm of clustering, topic modelling, or metric learning, can suffice.
Data visualization is a necessary step in the whole lifecycle of data mining. Good visualization techniques and systems facilitate pattern exploration. However, existing visualization techniques and solutions are general-purpose, and thus cannot suffice for certain research problem and analytical tasks. The changellen lies in two aspects. Firstly, novel and efficient visualization techniques is necessary. Secondly, well-designed visual analytics systems that facilitate patterns exploration at different granularities are desired.
Data visualization needs to be easy to understand and interpret for humans. By interacting with machines, humans receive the machine representable forms of information, get knowledge out of it, and take intutitive steps to refine the information. This process of crowd-sourcing involve human perception and expertise, such that machines can represent information in a more intutitive way.