Employing Machine Learning to Predict Student Aviator Performance

Abstract

Machine learning analysis of student aviator training performance data offers novel and more accurate methodologies for performance assessment that includes identifying students for attrition or remediation as well as optimal pipeline assignments. Machine learning provides an opportunity to better evaluate students by fully examining every indicator of performance throughout a student’s training: from subtest scores on the aviation selection test battery to test scores during initial ground school through each graded item on every flight event. The goal is to reduce time-to-train, improve aviator quality, and reduce training costs from failure to complete training. In this research, we identify important predictors and develop prediction models of performance in Primary, Intermediate, and Advanced flight training based on data from Aviation Selection Test Battery (ASTB), Introductory Flight Screening (IFS), and Aviation Preflight Indoctrination (API) training.