基于心脑血管疾病死亡数据的ARIMA模型、指数平滑法模型构建及其预测效能比较

Construction of ARIMA and exponential smoothing models based on cardiovascular and cerebrovascular disease mortality data and comparison of their predictive efficacy

目的基于心脑血管疾病死亡数据构建差分自回归移动平均模型(ARIMA模型)及指数平滑法模型,对比二者对疾病死亡预测效能,选出预测模型,为心脑血管疾病的防控工作提供科学依据。方法基于2017-2021年中国疾病预防控制信息系统人口死亡信息登记管理系统中的黑龙江省佳木斯市居民心脑血管疾病死亡数据,构建ARIMA模型与指数平滑法模型,预测2022年1-12月心脑血管疾病死亡人数,并与2022年1-12月实际值进行对比,以评估2种模型的预测效能。结果构建心脑血管疾病ARIMA模型的最优模型为ARIMA(0,1,1)(1,0,0),模型预测参数为平均绝对百分比误差(MAPE)=8.37%、贝叶斯信息准则(BIC)=8.696、均方根误差(RMSE)=69.722、平均绝对误差(MAE)=53.143。构建心脑血管疾病指数平滑法模型中的Holt-Winters可加性模型为最优模型,模型预测参数为MAPE=6.91%、BIC=8.200、RMSE=54.462、MAE=539.722。指数平滑法模型的拟合预测效果的平均绝对百分比误差小于ARIMA模型。结论基于心脑血管疾病死亡数据成功构建了ARIMA模型及指数平滑法模型。指数平滑法模型对心脑血管疾病死亡的预测效果优于ARIMA模型,适用于心脑血管疾病死亡情况的短期预测。

Objective Based on the death data of cardiovascular and cerebrovascular diseases,the differential autoregressive moving average model(ARIMA model)and the exponential smoothing model were constructed to compare the prediction efficiency of the two models for disease death,and the prediction model was selected to provide scientific basis for the prevention and control of cardiovascular and cerebrovascular diseases.Methods The data of deaths from cardiovascular and cerebrovascular diseases of residents in Jiamusi city from 2017 to 2021 were collected from Population Death Information Registration Management System of China Disease Control and Prevention Information System,and the ARIMA model and exponential smoothing model were constructed to predict the number of deaths from cardiovascular and cerebrovascular diseases from January to December 2022,and to compare it with the actual values in January-December in 2022,in order to assess the predictive efficacy of the 2 models.Results From 2017 to 2021,the number of deaths from cardiovascular and cerebrovascular diseases of Jiamusi residents was 44230,and the mortality rate was 381.94/105,showing an increasing trend.The optimal model for constructing ARIMA model of cardiovascular and cerebrovascular diseases was:ARIMA(0,1,1)(1,0,0),The prediction parameters of the model were mean absolute percentage error(MAPE)=8.37%,bayesian information criterion(BIC)=8.696,root mean square error(RMSE)=69.722,mean absolute error(MAE)=53.143.The Holt-Winters additivity model of the exponential smoothing method was the optimal model.The predicted parameters of the model were MAPE=6.91%,BIC=8.200,RMSE=54.462,MAE=539.722.The average absolute percentage error of the exponential smoothing model was smaller than that of the ARIMA model.Conclusion Based on the data of cardiovascular and cerebrovascular disease deaths,ARIMA model and exponential smoothing method model are successfully constructed.The exponential smoothing method model is more effective than the ARIMA model in predicting cardiovascular disease deaths and is suitable for short-term prediction of cardiovascular disease deaths.