摘要
目的建立基于Elman神经网络的感染性腹泻病人中细菌性食源性疾病阳性检出率预测模型，评估探讨El—man神经网络模型在细菌性食源性疾病发病预测中的应用价值。方法利用深圳市2008年1月至2012年12月的细菌性食源性疾病疫情资料作为训练集，建立Elman神经网络模型；选取深圳市2013年1—6月的细菌性食源性疾病资料作为检验集，评价该模型的预测效能。结果当网络结构为12—32．1．1时，构建的Elman回归网络模型为最优预测模型，此时训练集模拟仿真结果的平均误差均方为65．75。在此最优网络预测模型下，检验集预测值的平均误差绝对值为1．20，平均误差绝对率为0．21，非线性相关系数为0．79。结论基于Elman回归网络预测模型对细菌性食源性疾病发病具有较好的预测效能。
Objective To establish the incidence prediction model of bacterial foodborne diseases based on the Elman neural network, and explore the potential application of Elman neural network model to predict the rate of bacterial foodborne diseases in infectious diarrhea eases. Methods Elman neural network model was established using the epidemic data of baterieal foodborne disease in Shenzhen city from January 1,2008 to December 31,2012, and the predictive performance was measured and accessed using the data from January 1 to June 30,2013. Results The optimal Elman recurrent network structure was 12-32-1-1 ,and the simulation results mean error square （MSE） was 65.75. Further research found that the mean absolute error （MAE） of prediction results of the test-set using the optimal Elman neural network model was 1.20,the mean error rate （MER） was 0. 21, and nonlinear correlation coefficient （RNL） was 0.79. Conclusion The prediction model based on the Elman recurrent network was performed well on the incidence prediction of bacterial foodbome diseases.
出处
《公共卫生与预防医学》
2013年第6期10-13,共4页
Journal of Public Health and Preventive Medicine
基金
2011年深圳市科技局资助项目(NO.201102109)
