Accession Number : AD0686809

Title :   INVESTIGATION OF CAPER (COMPUTER AIDED PATTERN EVALUATION AND RECOGNITION) TECHNIQUES FOR PERSONNEL CLASSIFICATION.

Descriptive Note : Summary rept. 14 Jan 67-14 Feb 69,

Corporate Author : LOCKHEED MISSILES AND SPACE CO SUNNYVALE CALIF

Personal Author(s) : Mangelsdorf,John E. ; Bjorn,Theodore , Jr. ; Specht,Donald F.

Report Date : FEB 1969

Pagination or Media Count : 74

Abstract : The purpose of this research was to continue to investigate the applicability of nonlinear statistical pattern classification techniques to commonly encountered problems in the field of personnel research and personnel classification. The methods were primarily those developed by D. F. Specht and are referred to as CAPER (Computer) Aided Pattern Evaluation and Recognition) techniques. The primary technique used in the present study was NONLIN-ITLIN. This algorithm consists of a nonlinear transformation (NONLIN) followed by application of an iterative linear regression (ITLIN) solution. Major effort was spent in using Strong Vocational Interest Blank (SVIB) items to predict officer retention in the Naval service. The government-furnished data base consisted of the answers given by 1045 officers to the SVIB, plus the criterion information as to whether they remained in or separated from the service. The highest accuracy obtained was approximately 82 percent correct when 39 SVIB items were used as predictors. This result was obtained for a selection ratio of one. It is pointed out that the percent of officer retentees correctly identified could be increased to nearly 100 percent by reducing the selection ratio. NONLIN-ITLIN results were found superior to alternate methods studied. (Author)

Descriptors :   (*NAVAL PERSONNEL, CLASSIFICATION), (*PERSONNEL MANAGEMENT, PATTERN RECOGNITION), SELECTION, STATISTICAL ANALYSIS, ALGORITHMS, ITERATIONS, OFFICER PERSONNEL, QUESTIONNAIRES, COMPUTER PROGRAMMING, MATHEMATICAL PREDICTION

Subject Categories : Personnel Management and Labor Relations

Distribution Statement : APPROVED FOR PUBLIC RELEASE