Analysis of regional effects on market segment production

Abstract

This thesis develops a data-driven statistical model capable of identifying regional factors that affect the number of United States Army Recruiting Command (USAREC) accessions in Potential Rating Index Zip Code Market New Evolution (PRIZM NE) market segments. This model will aid USAREC G2 analysts involved in conducting recruiting market intelligence. Market intelligence helps the commander visualize the performance of subordinate units within their market and provides recommendations for use and expansion. This thesis first attempts to establish that a single high-assessing PRIZM NE market segment, Segment 32, does not access recruits at the same rate across regions. This thesis then develops general linear regression and gradient boosted decision tree models to determine the regional factors that contribute to the variance of recruit production. In particular, the gradient boosted decision tree delivers predictive results that allow analysts to identify regions that have underperforming accession rates compared to the national average. The recommendation of this thesis is that the USAREC implement the gradient boosted decision trees for use in G2 market analysis.