The goal of this project is to examine intelligence analysts’ workflow and build recommendation tools that can help analysts when examining a novel situation. Intelligence analysts are often overloaded with the amount of information that they have available to them, and they spend a substantial amount of time sorting through the information to identify the right information for a particular problem, and in some cases the right information might not even exist. We will begin the development of this tool based on a dataset that was created by the University of Kentucky. This dataset represents a detailed log of about 8166 actions taken by 25 intelligence analysts who used a handful of simple tools to investigate a fictional scenario and uncover a crime. These actions include things like “Read”, “Search”, “Switch”, as well as identifying the suspect. The goal of this project is to look at these logs, and build a tool that can identify which patterns of analyst activity were successful and which were not. Using this information the next step would be to build a recommendation engine for novel workflows that the analyst has not seen before. We plan to employ functional data analysis methods to model these patterns. The project research questions revolve around how to model this data in a way that helps identify patterns of successful users, and how to use that model to then make recommendations to analysts in a novel workflow.
Qualifications: Applicants must be currently enrolled in our PhD program and have proficiency in programming using R. Preferred candidates will possess expertise in multivariate data analysis and longitudinal data analysis.
If you are interested in this opportunity please email Ana-Maria Staicu at astaicu@ncsu.edu with your CV and the name of 1-2 NCSU professors who can speak about you. Applications will be considered on a first-come, first-served basis until the position is filled.