Risk analysis and monitoring of continuously arriving data are required by a wide range of domains critical to our individual and societal well-being, from the military to the financial sector, from medicine to homeland security, from man-made to natural disasters. These application areas use patterns or models to differentiate behavior of varying degrees of importance or risk, both in terms of the value for a particular variable, relationships between different variables, and changes across time. The objective of this project is the design, development, and assessment of visual analytics technology to support the real-time interactive analysis of data streams that focuses on creating and using models of stream behavior to identify instances of potentially important activities and risks in the data.
The research breaks new ground in the design of innovative exploration techniques of high-volume streaming data, with a focus on visualization and interactions of competing and complementary models extracted from the stream data and of change descriptions derived from each. Novel technology includes computational and visual methods for model formation, management, and model change analysis. Efficient indexing and compression of the history of how data and models change provide analysts with the ability to compare and contrast similar events over long periods of time. Linked views of the different information spaces (data, model, change, and history), and powerful interaction tools to support exploration and model building enable analysts to confirm the expected and uncover the unexpected.
The results of this project are expected to have significant impact is a wide range of domains that rely on extracting information from real-time digital data, including finance, medicine, and homeland security. In particular, tools are developed for solving financial risk and fraud detection problems critical to the economic well-being of individuals and organizations. Educational material from using the technology in courses are freely available, along with source code and data sets, via the project web page (http://davis.wpi.edu/~xmdv).
