论文部分内容阅读
目的对某院2011年1月-2016年5月乳腺外科出院人数进行时间序列预测分析,旨在为医院信息化、科学化管理提供参考依据,同时对ARIMA模型和Holt-Winters模型进行分析比较。方法通过ARIMA模型和Holt-Winters模型对乳腺外科出院人数进行模型训练,然后利用2016年6月、7月出院人数进行验证,并利用所建立的模型对2016年8月-12月出院人数进行预测。通过ARIMA模型和Holt-Winters模型对出院人数数据进行拟合,并与2016年6月和7月实际值进行比较。结果 ARIMA(0,1,2)(1,1,0)_(12)模型中非季节性移动平均系数MA1=-0.9549,季节性自回归系数SAR1=-0.3278。Box-Ljung统计量检验均无统计学意义(P=0.148),因此残差序列为白噪声序列。Holt-Winters模型对时间序列数据进行对数变换修正后的α,β,γ的估计值分别为0.19,0.31,0.76。Box-Ljung检验统计量无统计学意义(P=0.88),意味着延迟1~20阶为非零自相关,并且预测误差在整个时间段呈现出服从零均值,方差不变的正态分布。结论乳腺外科出院人数存在逐年上升的趋势,并且具有很明显的季节性趋势,Holt-Winters模型相比ARMIA模型对于短期的出院人数预测具有更强的实用价值,可作为科学管理提供参考依据。
Objective To predict the number of breast surgeries discharged from January 2011 to May 2016 in a hospital, and provide a reference for hospital informatization and scientific management. At the same time, the ARIMA model and Holt-Winters model are analyzed and compared. Methods The numbers of discharges of breast surgery were trained by ARIMA model and Holt-Winters model. The number of discharged patients in June and July 2016 was used for validation. The model was used to predict the number of discharged patients from August to December in 2016. . Discharged population data were fit using the ARIMA model and Holt-Winters model and compared with actual June and July 2016 values. Results The non-seasonal moving average coefficient MA1 = -0.9549 and the seasonal autoregressive coefficient SAR1 = -0.3278 in the ARIMA (0,1,2) (1,1,0) _ (12) model. Box-Ljung statistic test was not statistically significant (P = 0.148), so the residual sequence is a white noise sequence. Holt-Winters model of the time series data logarithmic transformation corrected α, β, γ estimates were 0.19,0.31,0.76. The Box-Ljung test statistic was not statistically significant (P = 0.88), implying delays of 1 to 20 were non-zero autocorrelations, and the prediction error presented a normal distribution obeying zero mean and variance over the entire time period. Conclusion The number of breast surgery discharges tends to increase year by year and has obvious seasonal trends. Compared with the ARMIA model, the Holt-Winters model has more practical value for short-term discharge number predictions and can be used as a reference for scientific management.