Spatio-Temporal Tensor Factorization (STTF) Approaches to Computational Phenotyping of Human Behavior
Abstract
In the coming years, digital phenotyping is likely to lead to a new science of behavior, going far beyond what is achievable by current practices. A phenotype is a concise, observable and understandable concept or pattern that is clinically meaningful, representing the essence of a particular medical condition or behavior, and actionable by practitioners. In this project, we propose to explore Spatio-Temporal Tensor Factorization (STTF) approaches to characterizingand predicting human behavior, with a particular eye towards characterizing those behaviors deemed unproductive or undesirable. This effort will entail advanced Machine Learning and other AI techniques, as well as the use of domain knowledge to constrain or refine obtained solutions.We will leverage our expertise in computational phenotyping, in addition to our close relationship with and understanding of the needs of the Navy, in order to build next-generation methods for behavioral phenotyping and explore their use in downstream applications. Our primary motivating application is the current effort by ONR and the Navy to create a Commander~s Risk MitigationDashboard to maintain situational awareness of the fitness and overall well-being of Navy units over time. This ongoing effort involves considerable investments in gathering data from multiple sources, and organizing and presenting them in the form of succinct descriptive statistics, for whichour proposed work will set the stage for advanced, prescriptive solutions.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 20, 2019
- Source ID
- N000141912625
Entities
People
- Ravi Srinivasan
Organizations
- Office of Naval Research
- United States Navy
- University of Texas at Austin