员工流动季节性问题及SARIMA-ANN模型应用

SARIMA-ANN Model Application in Seasonal Problem of Employee Turnover

中国企业员工流动受到中国商品生产季节性、劳动力市场供给周期性、员工身份两栖化等多种因素的影响,其数据呈现出季节性、非线性等特征。单一的自回归单整移动平均模型(ARIMA模型)不能较好地拟合其发展趋势并预测未来。本文通过神经网络模型(ANN模型)来修正传统的自回归单整移动平均模型(ARIMA模型),并加入季节性因素,从而形成SARIMA-ANN耦合模型,对企业员工流动的数据进行拟合与预测。通过对多组SARIMA-ANN模型的构建、衡量、比较与讨论,最终确定了较佳的神经网络对时间序列模型进行修正的耦合模型。实证结果显示,SARIMA-ANN模型充分考虑数据的季节性与趋势性,随机性与非线性特征并存的问题,对于季节性时间序列的经济数据的处理与预测是切实可行的。该耦合模型的应用证实了中国企业员工流动数据的趋势性与季节性、线性与非线性特征并存。这说明中国企业员工的流动具有更复杂的不规则的运动与突变,在精确预测有一定难度的情况下做好现有员工的留存工作是首要之策。

The data and its extension tendency of employee turnover in Chinese enterprises cannot be well fitted and predicted by a single Autoregressive Integrated Moving-average Model(ARIMA Model)because of the seasonality and non-linearity of the data.In this paper,the SARIMA-ANN coupled model formed through the traditional ARIMA Model is modified by artificial neural network model(ANN).And seasonal factors and the data and its extension tendency of employee turnover are fitted and predicted by the SARIMA-ANN coupled model.A case study with the data of 93 months in a Chinese matured enterprise is used to validate the coupled model.The best coupling model is finally determined through the construction,measurement,comparison and discussion of multi-group SAARIMA-ANNA models.The results of the case study show that SARIMA-ANN model is feasible to fit and predict economic data of seasonal time series because the model fully deal with the coexistence of the seasonality and tendency,randomness and non-linearity of data.The application of the coupling model verifies the existence of this feature,that is,the data of employee turnover in Chinese enterprises coexist with the tendency and seasonality,linearity and non-linearity.This indicates that the employee turnover in Chinese enterprises has more complicated irregular movement and mutation,and it is the first step to do a good job in retaining existing employees when it is difficult to accurately predict.