Accession Number : ADA443651
Title : Recruiter Selection Model
Descriptive Note : Technical rept. Dec 2005-Mar 2006
Corporate Author : MILITARY ACADEMY WEST POINT NY OPERATIONS RESEARCH CENTER
Personal Author(s) : Halstead, John B.
Full Text : http://www.dtic.mil/get-tr-doc/pdf?AD=ADA443651
Report Date : MAY 2006
Pagination or Media Count : 77
Abstract : This research enhances Industrial and Organizational Psychology (IO) by providing a statistical prediction of job performance derived from psychological inventories and biographical data. The research uses a combination of statistical learning, feature selection methods, and multivariate statistics to determine the better prediction function approximation with features obtained from the Non Commissioned Officer Leadership Skills Inventory (NLSI) and biographical data. The research created a methodology for iteratively developing a statistical learning model. After exploring RandomForest, Support Vector Regression, and Linear Regression, the RandomForest model best predicts recruiter performance for these data. The model's performance was further enhanced by using a greedy feature selection method to determine the best subset of features that produced the best model generalization. The resulting model runs in R statistical language and is controlled within an Excel worksheet environment by using Visual Basic Application (VBA) language to call R. The end product enables general user utilization of a statistically eloquent model, normally reserved for advanced researchers, engineers, statisticians, and economists. The model represents a multi-modal relationship primarily between recruiter age and NLSI score and to a lesser degree, 34 other features. This study is a result of Army Recruiting Initiatives earlier research into the feasibility of constructing a recruiter prediction model. The model, because of Excel deployment, is convenient to use. The convenience facilitates the model's migration from Center One to, perhaps, TRADOC. The model also provides significant cost benefits to the Army. If the model is used, recruiting potentially receives those individuals with the inherent skill sets for recruiting. Equipping recruiting with the right individual should reduce the number of required recruiters because the command's gross write rate should increase.
Descriptors : *MODELS , *PERFORMANCE(HUMAN) , *RECRUITERS , JOBS , STATISTICS , SELECTION , INDUSTRIAL PSYCHOLOGY , ARMY , MATHEMATICAL PREDICTION , RECRUITING , SKILLS
Subject Categories : PSYCHOLOGY
PERSONNEL MANAGEMENT AND LABOR RELATIONS
STATISTICS AND PROBABILITY
MILITARY FORCES AND ORGANIZATIONS
Distribution Statement : APPROVED FOR PUBLIC RELEASE