Collaborative Research: Non- and Semi-Parametric Modeling of Structured Human Activity Patterns Using Point Processes

Suny At Binghamton, Binghamton NY

Investigators

Abstract

This research project will advance theory and methods for analyzing structured human-activity data. Because of recent technological advances, large amounts of real-time human-activity data, such as social media data and transaction data, are being collected by various sources. The complexity and magnitude of these new data call for new statistical modeling tools. The methods to be developed in this project are motivated by and can be used to further understanding of human behaviors. Through active interdisciplinary collaborations with domain experts, the project will establish a bridge between the statistics community and the behavioral finance and social science communities. Open-source R packages will be developed and made available for public use through CRAN. Key materials from the project will be incorporated into the advanced graduate student courses. The investigators will develop a series of new non- and semi-parametric point process models for temporal point patterns of structured human activities. In particular, they will develop (1) a multi-level functional principal component analysis framework for modeling human activities such as stock trading etc.; (2) a simultaneous modeling and clustering approach for daily human activity patterns that are not only structured but also heterogeneous across different sub-populations; and (3) a new class of bivariate point process models to model the complex behaviors of modern social media users and provide meaningful insights into a user's content generating behavior. For the first two aims, popular functional data analysis tools will be introduced to model point processes with complex structures. Efficient computational algorithms will be developed and theoretical properties of these algorithms will be investigated. For the third aim, a semi-parametric regime-switching multi-type point process model will be used to model social media posting behaviors, where the posting intensity functions are approximated with spline basis functions. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.