隨著大數據的崛起，一些標竿企業人力資源管理部門開始運用Auto ML平台並導入機器學習，應用演算法建立模型，進行員工離職率管理。本研究以台灣C公司在大陸江蘇廠的員工資料進行分析，利用梯度提升、極限梯度提升、隨機森林這三種監督式機器學習分類演算法，建立自願性離職員工預測模型，並以交叉驗證方式處理，結果顯示，梯度提升及極限梯度提升兩個模型的AUC值均達0.8以上，代表模型具有優良的鑑別力，能有效預測員工是否具有離職傾向。 同時探討多項變數中影響員工離職之關鍵影響因素，及離職高風險群特徵。在18項自變數中找到關鍵影響因素，分別為職等為6、績效平均在2.83分以下、年資在11.4年以下、非主管職、職稱為工程師或管理師、及戶籍為江蘇省共七項，當中最有可能離職為職等6員工族群，次高為員工績效平均分數在2.83分以下。此外，研究結果顯示高風險群特徵為職等6、住在當地且大學畢業者。綜合歸納本次研究分析結果，提供C公司員工離職率管理之參考。 ;With the rise of big data, the human resource management department of some benchmark enterprises have started to use Auto ML platform and introduced machine learning to build models by applying algorithms for employee turnover rate management. This study analyzed the employee data of Jiangsu plant in China of Taiwanese Company C. Using three supervised machine learning classification algorithms, namely, gradient boosting, extreme gradient boosting (XGBoost) and random forest to build a voluntary turnover prediction model, and processed it by cross-validation. The results showed that the AUC of both gradient boosting and extreme gradient boosting (XGBoost) models were above 0.8, indicating that the models had good discriminative ability and could effectively predict whether employees have a tendency to leave. The key influencing factors of multi-variables that affect employee turnover and the characteristics of the high-risk group were also investigated. Between 18 independent variables, 7 key factors were found: being in level 6 position, getting the average performance score of 2.83 or less, having 11.4 or less years of experience, being in non-supervisory position, holding a posititon title of engineer or administrator, and household being registered in Jiangsu Province. Among them, the most likely to leave is the employee who is in level 6 position, and the second highest is the employee with an average performance score of 2.83 or below. In addition, the results showed that the characteristics of the high-risk group are the ones who are in level 6 position, live in the local area and are graduated from university. The results of this study are summarized to provide Company C a reference for employee turnover rate management.
