關鍵詞：機器學習、人員自願性離職、特徵選取、預測模型、監督式分類 ;This study collects data from a manufacturing company in Taiwan. Logistic regression, support vector machine, decision tree, random forest, and eXtreme Gradient Boosting algorithms are adopted in order to build a reliable predictive model to predict employee voluntary turnover. Moreover, imbalanced classification problem, feature selection and K-fold cross validation are introduced and tested in this study. The results suggest random forest and eXtreme Gradient Boosting perform the best, both predictive models have the F-Score and AUC values above 0.85. Results of variable importance show elementary level of managerial training hours, professional training hours, average number of promotions, job tenure, and age contribute the most in predicting employee voluntary turnover outcome.
