Utilization Of Machine Learning Techniques To Detect Anomalies In Department Of Defense Contract Data
Abstract
The Federal Funding Accountability and Transparency Act (FFATA) of 2006 makes available to public view information on all federal contracts beginning in 2006. This transparency presents an opportunity to examine large volumes of procurement data, in particular to infer whether anomalies or irregularities are present. In this thesis, we examine direct-order purchases made by the U.S. Army between calendar years 2013 and 2017. A total of 73,570 direct-order contracts were issued by the Army during this period, with a total obligation value of over $36 billion. We use supervised machine learning techniques to detect trends regarding levels of competition, set-aside programs used, sole sourcing, and monies spent both in individual contracts and in the awarding offices that issued the contracts. We also identify specific contracts that warrant further inspection. The suite of analytical tools that we develop can be applied generally to direct-order contracts issued by other DoD service branches. Application of these tools would allow an investigator to identify DoD contracts that warrant further scrutiny, and would allow contracting activities to be monitored with respect to criteria that are identified with best spending practices.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2019
- Accession Number
- AD1073639
Entities
People
- Nicholas L. Lee
Organizations
- Naval Postgraduate School