Theoretical and Applied Data Science Lunch-n-learn: Two Challenge Problems from the Algorithms for Threat Detection Program

Thursday, October 31, 2019 - 12:00pm to 1:00pm
Event Type: 

Presenter: Lee Przybylski and Nate Harding

Title: Two Challenge Problems from the Algorithms for Threat Detection Program

Abstract:

As part of the Algorithms for Threat Detection program, our team participated in two challenge problems.  The first was an association problem.  Given a list of AIS vessel tracking data, our objective was to identify which points belong to the same vessel.  Our approach was to design an estimator on the coupling between data points and apply a greedy forensic algorithm.  The second problem involved detecting traffic incidents.  Our objective was to predict the times and locations of traffic incidents based on the volume of traffic.  To detect the anomalies, we applied topological data analysis in the form of persistence diagrams.  We then used a bagging method to measure the disparity between an unbiased approximation of normal traffic behavior and abnormal behavior.

Bio:

Lee Przybylski is a 3rd year PhD student in Mathematics at Iowa State University. 

Nate Harding is a 4th year PhD student in Mathematics at Iowa State University.

Category: