摘要
目的比较季节性差分自回归滑动平均模型(SARIMA)与SARIMA-广义回归神经网络(GRNN)组合模型在肺结核发病预测中的应用,探讨更为准确的模型为完善肺结核预测预警系统提供参考。方法研究纳入了2004-2017年间昆山市结核病月度发病数据,采用时间序列分析中的SARIMA和GRNN技术。2004-2016年间的数据作为建模数据,2017年的数据作为验证数据。结果 2004-2017年间,昆山市肺结核发病率呈现缓慢下降趋势,平均每年下降2. 51%,且存在明显的季节性规律,冬春季较多。SARIMA (0,1,1)(0,1,1)_(12)模型较好的拟合了昆山市结核病发病长期趋势和季节性,其平均误差率为20. 66%,决定系数为0. 799。SARIMA (0,1,1)(0,1,1)_(12)–GNRR组合模型的评价误差率为13. 83%,决定系数为0. 833。SARIMA (0,1,1)(0,1,1)_(12)–GNRR组合模型较SARIMA(0,1,1)(0,1,1)_(12)模型的优越性在2017年度的验证数据中进步一得到证实。结论近年来肺结核发病逐渐下降;结核病的发病具有季节性,高峰集中在冬春季; SARIMA (0,1,1)(0,1,1)_(12)–GNRR组合模型较SARIMA(0,1,1)(0,1,1)_(12)模型能更好的拟合结核病发病长期趋势和季节性。
Objective This study aimed to comparing the application of SARIMA and SARIMA-GRNN in predicting the incidence of tuberculosis. To explore more accurate models to provide reference for perfecting tuberculosis prediction and early warning system. Methods This study included the monthly incidence of tuberculosis in Kunshan from 2004 to 2017. Data from 2004 to 2016 was used as modeling data,and data from2017 was used as verification data. Results The incidence of tuberculosis in our city showed a slow downward trend,with an average annual decline of 2. 51%,and there were obvious seasonal patterns,more in winter and spring. SARIMA( 0,1,1)( 0,1,1)12 was identified. The mean error rate of the single SARIMA model and the SARIMA-GRNN combination model was 20. 66% and 13. 83%,and the determination coefficient was 0. 799 and 0. 833 respectively. The better performance of the SARIMA-GRNN combination model was further confirmed with the forecasting data set( 2017).Conclusion Tuberculosis was a seasonal disease,with a predominant peak in spring and winter. The SARIMA-GRNN model was more effective than the widely used SARIMA model.
作者
王华
田昌伟
王文明
滕国兴
WANG Hua;TIAN Changwei;WANG Wenming;TENG Guoxing(Kunshan Center for Disease Control and Prevention , Soochow 215300, Jiangsu Province, China;School of Public Health , Soochow University, Soochow 215123 , Jiangsu Province, China)
出处
《寄生虫病与感染性疾病》
CAS
2019年第1期28-31,共4页
Parasitoses and Infectious Diseases
关键词
肺结核
季节性差分自回归滑动平均模型
广义回归神经网络
pulmonary tuberculosis
seasonal autoregressive integrated moving average model
generalized regression neural network model
