Aiming at the problem that employee turnover will increase the operating cost and reduce the profitability of the enterprise, this paper proposes the prediction algorithm of the resigned employees using machine learning algorithm. The LRA prediction model is constructed by combining Adaboost and Random Forest basic algorithms with Stacking ensemble algorithm to predict employee turnover in an enterprise. Experimental results show that the prediction accuracy of LRA model is 89.09%, which is significantly higher than that of the model constructed by a single algorithm. Precision, recall and F1 metrics confirm the feasibility and reliability of the model.By sorting the importance of the input LRA model’s features, the main factors affecting employee resignation are overtime, length of service (0-3 years), income, occupational level, etc., this method enriches the conclusions of existing studies and is beneficial to enterprise decision-makers to make reasonable decisions on employee turnover behavior.