半導體測試業之員工離職預測研究－以K公司為例

Abstract

人力資源是對企業最重要的資產，人力的組成結構將影響企業的相對競爭力，經過2008年的全球性不景氣，企業為了降低成本進行人力的重整，因此可能造成員工的向心力和投入程度大幅下降。 除了如何在不景氣時保留或培養未來所需要的人力，為即將到來的回春景氣做準備是對企業重要課題，如何預測人員異動的狀態，進行離職的預測及預防，並及時採取優秀人才的留才行動是重要的。 本研究針對企業之人力資源管理系統資料，包含人員基本資料、考勤資料、薪資資料及晉升異動資料進行基本的分類及探勘作業。分類是透過資料屬性特徵相似度進行分類；探勘機制則是對以上四種已整理的資料進行探勘分析，使用演算法包含「決策樹演算法」及「關聯規則演算法」。 實驗結果顯示，「關聯規則演算法」的預測結果準確機機率較高，對於半導體測試業資料模型之員工離職預測有較好的表現。

Human Resources are the most important asset of the enterprise, and the Manpower composition will affect the relative competitiveness of enterprises. After the global recession in 2008, the manpower restructuring for reducing the cost may cause the centripetal force and commitment of employees substantially decline. Turnover implies the endeavor to find another job. For companies, it means an increase in personnel expense. So, it is very important that how to predict employee turnover intention, and execute the policies for resignation inclining forecast and brain drain prevention. This study uses human resource management system, including personnel characteristics, attendance status information and the promotion and salary information, for the basic classification and data mining operations. Decision tree algorithm and Association rule algorithm are adopted to build the prediction models for employee Turnover prediction. The experimental results show that the association rules approach can have higher accuracy of the predictions.