自回归求和移动平均乘积季节模型在我国布鲁菌病短期月发病人数预测中的应用

发布时间：2019-04-08 18:29

【摘要】：目的研究我国布鲁菌病(布病)月发病人数的趋势性和季节性,探讨自回归求和移动平均(ARIMA)乘积季节模型预测我国布病短期月发病人数的效果。方法收集2004年1月—2015年5月我国布病月发病人数(共137组),进行时间序列分析。数据来自国家卫生和计划生育委员会公布的疫情监测数据。观察我国布病月发病人数的趋势性和季节性,以我国2004—2013年的布病月发病人数作为训练样本,拟合ARIMA乘积季节模型;用2014年1月—2015年5月的发病数据作为校验样本,验证模型;确定最优模型后,预测2015年6—12月我国布病月发病人数。结果 2004—2008年我国布病月发病人数相对平稳,从2009年以后有了明显的上升趋势。从季节性来看,每年的6、7、8月属高发病期,每年的1月和12月处于全年的最低发病期。选取的最优模型为ARIMA(0,1,0)(1,1,0)12,其平均绝对百分误差(MAPE)=13.60,决定系数(R2)=0.881;对模型进行参数显著性检验,一阶季节自回归项(SAR)参数估计值=-0.292,P=0.048。运用ARIMA(0,1,0)(1,1,0)12对2015年6—12月我国布病月发病人数进行预测,其预测值分别为7 709、7 524、6 113、4 458、3 450、3 576、3 760例。结论从2009年以后,我国布病月发病人数有明显的上升趋势;季节性表现在6~8月为高发病期,12月至来年1月为低发病期。ARIMA乘积季节模型拟合我国布病月发病人数的时间序列模型精度较高,可以用来预测我国布病短期月发病人数。
[Abstract]:Aim to study the trend and seasonality of monthly brucellosis (brucellosis) in China, and to explore the effect of predicting the short-term monthly brucellosis in China by using the seasonal model of autoregressive summation and moving average (ARIMA) product. Methods the monthly incidence of brucellosis in China from January 2004 to May 2015 (137 groups) was collected and the time series analysis was carried out. The data came from surveillance data released by the National Health and Family Planning Commission. The trend and seasonality of the monthly incidence of brucellosis in China were observed. The monthly number of brucellosis in China from 2004 to 2013 was used as training sample to fit the seasonal model of ARIMA product. The data from January 2014 to May 2015 were used to verify the model, and the optimal model was determined to predict the monthly incidence of brucellosis in China from June to December in 2015. Results the monthly incidence of brucellosis in China was relatively stable from 2004 to 2008, and there was an obvious upward trend since 2009. From the seasonal point of view, 6, 7 and 8 months of the year are high incidence period, each year in January and December is the lowest incidence period of the year. The optimal model is ARIMA (0, 1, 0) (1, 1, 0) 12, the average absolute percent error (MAPE) = 13.60 and the determination coefficient (R ~ 2) = 0.881; The parameter significance test of the model shows that the (SAR) parameter of the first order seasonal autoregressive term is estimated to be-0.292, P = 0.048. ARIMA (0, 1, 0) (1, 1, 0) 12 was used to predict the number of epidemic diseases in China from June to December in 2015. The predicted values were 7 709, 7 524, 6 113, 4 458, 3 450, 3 576 and 3 760 cases, respectively. the predicted values were 7 709, 7 524, 6 113, 4 458, 3 450, 3 576 and 3 760 respectively. Conclusion since 2009, the monthly incidence of brucellosis in China has an obvious upward trend. Seasonal manifestations were high incidence period from June to August and low incidence period from December to January of the following year. Arima product seasonal model had high accuracy in fitting the monthly incidence of brucellosis in China, which could be used to predict the short-term monthly incidence of brucellosis in China.
【作者单位】：兰州大学医学院公共卫生学院;西北民族大学西部环境健康研究所;兰州交通大学测绘与地理信息学院;
【基金】：国家科技惠民计划资助项目(2012GS620101)
【分类号】：R516.7

【参考文献】