流行性脑脊髓膜炎与气象因素关系的BP神经网络模型研究

The model of back-propagation neural network about meteorological factors and epidemic cerebrospinal meningitis

目的:探讨流行性脑脊髓膜炎(流脑)发病率与气象因素的关系,建立流脑发病率的BP神经网络预测模型,评价模型效果。方法:利用SPSS 10.0统计软件进行气象因素与流脑发病率的相关分析。利用M atlab 6.5软件构建流脑发病率的BP人工神经网络预测模型。结果:相关分析结果显示流脑的发病率与平均气压、平均降水量呈负相关,与平均蒸发量呈正相关。BP神经网络模型的拟合结果显示,流脑发病率回代值的MER=1.73%、R2=0.9900,模型拟合效果较好;模型的预测精度为5.88。结论:平均气压、平均蒸发量、平均降水量对流脑发病率影响较大。BP神经网络模型对流脑发病率具有较高的拟合和预测能力。

Objective. To investigate the relationship between meteorological factors and the incidence of eerebrospinal meningitis and to build and evaluate the back-propagation （BP） artificial neural network model. Methods：The data of the incidence of epidemic cercbrospinal meningitis and meteorological factors from 1981 to 1994 were collected and analyzed by using SPSS10.0. The BP artificial neural network model was built by using Matlab 6.5. Results：The incidence of epidemic cerebrospinal meningitis was negatively correlated to annual mean atmospheric pressure and annual mean precipitation and positively correlated to annual mean evaporation. The mean error rate（MER） and coefficient of determination （ R^2 ） of BP model were 1. 73% and 0.9900, respectively. The forecasting precision of BP modal was 5.88%. Conclusion：The incidence of cerebrospinal meningitis is correlated to atmospheric pressure, precipitation, and evaporation. The BP neural network model fits well in the study of respiratory infectious diseases.