GM-ARIMA模型在医院死亡人数预测中的应用

The Application of GM-ARIMA Model in the Prediction of Deaths Numbers of Hospital

目的通过ARIMA模型和GM(1,1)模型的建立,探索一个新的GM-ARIMA模型,用以进行医院死亡人数预测。方法采用某医院2006年-2015年10年间的死亡人数数据,分别建立ARIMA模型和GM(1,1)模型。通过对2种模型预测值图形比较,创建GM-ARIMA混合模型。运用MAPE评价3种模型的预测效果。结果 ARIMA模型的MAPE值为3.9%,GM(1,1)模型为4.0%,而GM-ARIMA模型仅为3.1%,误差率最小,与实际情况最为接近。所以选择GM-ARIMA模型对医院死亡人数进行预测。结论 3种模型都可应用于死亡人数预测中。但是在选择模型进行预测分析时,应考虑影响模型的多种可能性因素,在模型拟合时体现出来,根据特定时间的不同情况对模型进行不断修正,保证模型的有效性。

Objectives Based on ARIMA model and GM（1, 1） model, a new GM-ARIMA model was explored to predict the number of hospital deaths. Methods The deaths data of a hospital for 10 years from 2006 to 2015 was adopted, and then used to set up ARIMA model and GM（1, 1） model respectively. Compare with the two kinds of model prediction graphics to establish a new GM-ARIMA model. Use MAPE to evaluate the prediction effect of three kinds of models. Results MAPE Error rate of GM-ARIMA model was the smallest, which was most closely with the actual death. Conclusions Three models could be applied in predicting the number of deaths. However,when choosing model to forecast and analysis, we should consider the factors that influence the various possibilities of the model, which is proposed to be reflected. According to the different situation of a certain time, the model needed to be corrected unceasingly to ensure the validity of the model.