Abstract：Objective To evaluate the effectiveness of ARIMA and GM(1,1) models in outpatient population forecasting, and to provide the basis for selecting appropriate predictive methods. Methods ARIMA model and GM (1, 1) model were constructed using monthly data of outpatient visits from 2005 to 2014. At the same time, the number of outpatients in 2015 was forecasted, and the results of the two models were validated by the actual number of outpatients in 2015. The evaluation indexes were average relative error and mean absolute value error. Results The average error rate (MER) of ARIMA and GM(1,1) models were 3.90% and 4.41%, respectively. The coefficients of determination R2 were 0.961 and 0.955, respectively. For the year 2015, the average relative errors between the predicted value and the measured value were 23 959 and 35 397, the average absolute errors were 4.62% and 6.83% respectively. Conclusion ARIMA model has a better effect than GM (1,1) model on samples over 30 and on the long-term trend, cyclical trend and implicit seasonal change trend. For the data such as the number of outpatients in time series, we should consider ARIMA Model first to get a better fitting effect, and the relative error is smaller.
收稿日期: 2016-09-28
引用本文:
刘国柱. ARIMA模型与GM(1,1) 模型在医院门诊人次预测中的比较研究[J]. 中国医院统计, 2017, 24(1): 5-8.
Liu Guozhu. Comparative study of ARIMA model and GM (1,1) model in hospital outpatient service forecast. journal1, 2017, 24(1): 5-8.
