ARIMA乘积季节模型在出院人数预测中的应用

Application of multiple Seasonal Autoregressive Integrated Moving Average Model in Forecasting the Number of Discharged Patients

目的了解医院住院量的变动趋势,对医院出院人数进行预测分析,为科学决策提供依据。方法应用乘积季节ARIMA模型对某院2003年1月-2013年12月出院人数进行模型拟合,预测2014年各月出院人数,用2014年1月-6月份实际资料评估模型的预测效果。结果该院出院人数呈明显的季节效应,且出院人数逐年小幅递增;乘积季节ARIMA(1,1,1)×(0,1,1)12(不含常数项)模型为最优模型,标准化的BIC(标准化贝叶斯信息量)和平均绝对误差百分比(MAPE)值最小,BIC值为11.98,MAPE值为5.43。Ljung-Box检验无统计学意义(Q18=10.575,P=0.782)。结论乘积季节ARIMA模型可以较好地拟合出院人数的变化趋势,是一种短期预测精度较高的预测模型。

Objective To investigate inpatient quantity trend, forecast the number of discharged patients ,in order to provide basis for scientific decision.Methods ARIMA model was used to fit the number of discharged patients from January 2003 to December 2013 in the hospital by multiple seasonal autoregressive integrated moving average model,to predict the number of discharged patients from January to December 2014. The model was evaluated by actual data from January to June 2014. Results The seasonal effect in the number of discharged patients was observed in the hospital, and the incidence was slightly increased over time. Multiple seasonal（1,1,1,）×（0,1,1）ARIMA 12（have no constant） model has been found as the most suitable mode with least Normalized Bayesian Information Criteria（BIC）of 11.98 and Mean Absolute Percent Error（MAPE）of 5.43. The mode was further validated by Ljung？Box test（Q18=10.575,P=0.782）. Conclusion Multiple seasonal ARIMAmodel can be used to fit the changes of the number of discharged and it is a predicted model of high precision for short time forecast.