Advancements in telecommunications and high-speed networks have created a unique category of digital information known as data streams. This time-varying information has the unique characteristic of arriving continuously, unpredictably, and unboundedly without any persistent patterns. Data stream examples include social network communications, cyber transactions, telecommunications, newswires, and remote sensing or high speed camera imagery. The increasing demands of immediate analyses and actions on these transient data streams in many time-sensitive applications, such as e-commerce/e-government and information assurance/homeland security, have spawned a series of investigations to query, mine, and model the information through non-traditional approaches.

Researchers at PNNL have successfully developed a visual-based technology to the fast growing research area of transient data stream analytics. The underlying visual analytics algorithm is designed to 1) ingest stream information adaptively when influx rate exceeds processing rate, and 2) adaptively and incrementally project newly arriving data onto an existing visualization of historical information without re-computation continuously.

The technology has been successfully applied to analyze data streams such as remote sensing imagery and newswires. Recently, the technology has also been proposed to analyze large-scale cyber transactions on a high-performance computing machine for the government. The signature-based design of the technology can be applied broadly to data streams gathered from different domains and applications.