Computer scientists to study application of big data for improved disaster management and response
Wednesday, Jan. 10, 2018
MANHATTAN — With the continuing threat of deadly disasters such as tornadoes, hurricanes, earthquakes and terrorist attacks striking communities, researchers at Kansas State University are looking to big data for its potential to become an integral source of information for response organizations in these situations.
Funded by $900,000 from the National Science Foundation Critical Techniques, Technologies and Methodologies for Advancing Foundations and Applications of Big Data Sciences and Engineering program, three faculty members from the computer science department in the College of Engineering are set to explore this potential through the four-year project, "Domain Adaptation Approaches for Classifying Crisis-Related Data on Social Media."
"Big crisis data can help improve situational awareness and facilitate faster response where it's most needed," said Doina Caragea, associate professor and co-lead investigator on the project. "Manually sifting through voluminous streaming data to filter useful information in real time is inherently impossible.
"This project aims to explore domain adaptation solutions based on deep learning to help emergency response organizations deal with the overload of information in real time," she said.
The research, co-lead by Cornelia Caragea, associate professor, and co-investigated by Dan Andresen, professor, will produce a hybrid community-computational framework for real-time discovery of situational awareness information in social media platforms.
"This framework has the potential to transform the way in which response organizations operate by helping them identify more relevant information in a timely manner and in turn helping them provide better support to victims of disasters," Doina Caragea said.
Amazon Web Services provided an additional $768,916 in promotional credits for performing the large-scale computation involved with training neural network models, almost doubling the funding for the project.
"We will build our framework on Apache Spark, using machines equipped with graphics processing units available to us through Amazon Web Services," Doina Caragea said. "These resources will enable the analysis of increasingly overwhelming amounts of crisis data, which are directly contributed by people 'on the ground.'"
With a combined involvement of graduate and undergraduate students, the project will make significant advances to the current state of the art in machine learning and its applications to disaster management and response.