Zhang Ying, Shao Yi, Shang Kezheng, Wang Shigong, Wang Jinyan
Wei Sheng Yan Jiu. 2014 Sep;43(5):774-8.
Set up the model of forecasting the number of circulatorys death toll based on back-propagation (BP) artificial neural networks discuss the relationship between the circulatory system diseases death toll meteorological factors and ambient air pollution.
The data of tem deaths, meteorological factors, and ambient air pollution within the m 2004 to 2009 in Nanjing were collected. On the basis of analyzing the ficient between CSDDT meteorological factors and ambient air pollution, leutral network model of CSDDT was built for 2004 - 2008 based on factors and ambient air pollution within the same time, and the data of 2009 est the predictive power of the model.
There was a closely system diseases relationship between meteorological factors, ambient air pollution and the circulatory system diseases death toll. The ANN model structure was 17 -16 -1, 17 input notes, 16 hidden notes and 1 output note. The training precision was 0. 005 and the final error was 0. 004 999 42 after 487 training steps. The results of forecast show that predict accuracy over 78. 62%.
This method is easy to be finished with smaller error, and higher ability on circulatory system death toll on independent prediction, which can provide a new method for forecasting medical-meteorological forecast and have the value of further research.
建立基于反向传播（BP）人工神经网络的循环系统死亡人数预测模型，探讨循环系统疾病死亡人数与气象因素及环境空气污染之间的关系。
收集南京市2004年至2009年期间的死亡人数、气象因素及环境空气污染数据。在分析循环系统疾病死亡人数与气象因素及环境空气污染之间相关性的基础上，基于同一时期的气象因素和环境空气污染数据，建立2004 - 2008年循环系统疾病死亡人数的神经网络模型，并用2009年的数据检验该模型的预测能力。
气象因素、环境空气污染与循环系统疾病死亡人数之间存在密切关系。人工神经网络模型结构为17 - 16 - 1，即17个输入节点、16个隐藏节点和1个输出节点。训练精度为0.005，经过487步训练后最终误差为0.00499942。预测结果显示预测准确率超过78.62%。
该方法易于实现且误差较小，对循环系统死亡人数具有较高的独立预测能力，可为医学气象预测提供一种新方法，具有进一步研究的价值。
