论文部分内容阅读
目的:应用自回归求和移动平均(autoregressive integrated moving average,ARIMA)乘积季节模型预测重庆市流行性乙型脑炎(简称乙脑)发病数。方法:利用R软件对重庆市2006年1月到2015年6月乙脑报告病例数进行ARIMA模型建模拟合,选择预测模型进行相互比较确定最优模型。用2015年7至12月实际报告病例数与拟合值的比较来评价模型的预测效果,并对2016至2017年重庆市乙脑报告发病数进行预测。结果:重庆市乙脑发病人数呈逐年下降趋势,报告病例具有明显季节分布特征,ARIMA(0,0,1)×(1,1,2)12模型很好地拟合了时间序列,该模型赤池信息量准则(Akaike information criterion,AIC)、许瓦兹贝叶斯准则(Schwarz Bayesian criterion,SBC)值均最小且预测值与实际值的平均相对误差为0.12,平均绝对百分比误差为7.81%。进一步用该模型预测重庆市2016至2017年乙脑病例数分别为35例和32例,发病高峰仍是7至8月。结论:利用ARIMA乘积季节模型对乙脑发病数拟合较好,短期预测结果良好;与2015年比较,预测2016至2017年乙脑报告发病数略微减少。
Objective: To predict the incidence of Japanese encephalitis (JE) in Chongqing by autoregressive integrated moving average (ARIMA) product season model. Methods: The RIM software was used to model and simulate the reported cases of JE from January 2006 to June 2015 in Chongqing, and the prediction model was selected for comparison with each other to determine the optimal model. The predicted results of the model were compared with the actual reported cases and fitted values from July to December 2015, and the incidence of JE reported in Chongqing from 2016 to 2017 was predicted. Results: The incidence of Japanese encephalitis in Chongqing showed a declining trend year by year. The reported cases had obvious seasonal distribution. The ARIMA (0,0,1) × (1,1,2) 12 model fitted the time series well. The model Akaike information criterion (AIC) and Schwarz Bayesian criterion (SBC) values are the smallest and the average relative error between predicted and actual values is 0.12 and the average absolute percentage error is 7.81%. Further use of the model predicts that the number of JE cases in Chongqing from 2016 to 2017 are 35 cases and 32 cases, respectively. The peak incidence is still between July and August. Conclusions: The ARIMA product seasonal model fitted well to the incidence of Japanese encephalitis, and the short-term predictive result was good. Compared with 2015, the incidence of Japanese encephalitis in 2016-2017 was slightly reduced.