论文部分内容阅读
目的探讨循环系统疾病死亡人数与气象条件及污染物的关系,建立基于误差反传算法(back-propagation,BP)神经网络的循环系统疾病死亡人数预报模型。方法选取南京市2004~2009年循环系统疾病死亡病例和同期的气象资料及污染资料。在对循环系统疾病死亡人数与气象因子和污染物进行相关分析的基础上,利用2004~2008年的气象和污染数据建立循环系统疾病死亡人数的BP神经网络预报模型,并用2009年的资料对该模型进行预报效果检验。结果气象因子和污染物均与循环系统疾病死亡人数密切相关。建立的循环系统疾病死亡人数的神经网络模型结果为17-16-1(即有17个输入、16个隐含节点和1个输出),训练精度为0.005,训练了487步达到目的,最终误差为0.004 999 42,预测准确率达78.62%以上。结论该方法计算简便,误差较小,对循环系统疾病死亡人数有较好的预测效果,为医疗气象预报提供了一种新方法,具有进一步的研究价值。
Objective To investigate the relationship between the number of deaths caused by circulatory diseases and meteorological conditions and pollutants, and to establish a prediction model of death toll of circulatory diseases based on Back-propagation (BP) neural network. Methods The death cases of circulatory diseases and the meteorological data and pollution data of Nanjing from 2004 to 2009 were selected. Based on the correlation analysis between the death toll of circulatory system diseases and meteorological factors and pollutants, the BP neural network forecast model of the death toll of circulatory diseases was established by using the meteorological and pollution data from 2004 to 2008, Model for forecasting effect test. Results Both meteorological factors and pollutants were closely related to the number of deaths from circulatory diseases. The neural network model established for the death toll of circulatory diseases results in 17-16-1 (ie, 17 inputs, 16 implicit nodes and 1 output) with a training precision of 0.005, 487 training steps to achieve the goal, the final error 0.004 999 42, the forecast accuracy rate is over 78.62%. Conclusions This method is simple, less error-prone, and has a better predictive value for the death toll of circulatory diseases, providing a new method for medical meteorological forecasting and has further research value.