AAAI Publications, The Thirty-First International Flairs Conference

Font Size: 
Predictive Models of User Performance for Marksmanship Training
Mary Jean Blink, Ted Carmichael, Jennifer Murphy, Michael Eagle

Last modified: 2018-05-10

Abstract


How the Army conducts rifle marksmanship training is undergo-ing a number of positive changes. Despite this, challenges to con-ducting and coordinating this critical training remain. One chal-lenge to assessing training effectiveness is a lack of persistent records of soldier performance; too often soldier data are purged shortly after training events for convenience and in order to en-sure privacy. This paper reports on our efforts to research the fea-sibility of collecting, analyzing, and storing data from multiple training systems, in order to accelerate and improve marksman-ship training. We do this through the use of cognitive, psychomo-tor, and affective constructs; and the use of predictive modeling techniques in order to forecast marksmanship qualification scores.These models successfully predicted scores on a 40-point scalewith a root mean square error (RMSE) of less than three, using models that are robust to changing input variables. Future im-provements and directions for this research are also discussed.

Keywords


Marksmanship Training; Learner Modeling; Cognitive Modeling; Affective States; Educational Data Mining; Predictive Modeling

Full Text: PDF