데이터 마이닝을 이용한 출동대원 이직등급 예측 모형

Abstract

The purpose of this thesis is to identify suitable forecast model for patrol guard’s resignation using data mining techniques after comparing and analysing dataming techniques based on sample data for patrol guard of physical security company in Korea. SAS Enterprise Miner, data mining software, has been used to implement forecast model and three data mining techniques are decision tree, logistic regression and neural network. Assessment for model is firstly measured accurate rate of misclassification table as a percentage. And secondly, ROC(Receiver Operating Characteristic) curve curve assesses to identify the best forecast model. The result shows that logistic regression is the better model than decision tree model and neural network model in classification accuracy of misclassification table for patrol guard resignation forecast. It is expected that physical security company in Korea will be able to process manpower management more systematically and efficiency when security company manages patrol guards of high resignation level through logistic regression model that is finally chosen. Meanwhile, sample data for this thesis tends to be slanted toward logistic regression. And verification for suitable variable selection and threshold selection is another research subject.