人工神经网络在呼吸系统疾病急诊就诊人数预报中的应用

Application of an artificial neural network in forecasting the number of emergency patients of the respiratory system

利用北京市2009-2011年呼吸系统疾病急诊就诊人数资料和同期的气象资料及污染资料,分析了气象因素及污染物分别与上感、下感急诊就诊人数的相关性,在此基础上,通过BP人工神经网络分别建立了上感和下感急诊就诊人数的预报模型,并对其效果进行评价.结果表明:气象因素和污染物与上感、下感的发病有密切的关系;建立的上感、下感就诊急诊人数的神经网络预报模型结构分别为13-7-1(即有13个输入、7个隐含节点和1个输出)和13-6-1(即有13个输入、6个隐含节点和1个输出),预测准确率分别为77.11%和75.57%.与统计预报方法相比较,该方法计算简便、误差较小、预测准确率高,对上感和下感急诊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.

Using the data on the respiratory system emergency patients, meteorological factors data and major air pollutants data within the same time in Beijing from 2009 to 2011, the correlations between meteorological factors, ambient air pollution and upper （low） respiratory tract infections emergency patients （URTE, LRTE） were analyzed and, then, a back-propagation （BP） artificial neutral network （ANN） model was built and eval-uated. The result showed that a close relationship existed between the meteorological factors, three pollutants and both URTE and IRTE. The ANN prediction model structure was 13-7-1 for URTE and 13-6-1 for IRTE;that is to say, there were 13 input notes, 7 hidden notes and 1 output note for URTE model, which was similar to the IRTE model. The results of forecast showed that the predicting accuracy were 77.11% for URTE and 75.57% for IRTE. Compared with traditional statistical forecasting methods, this method is easy to conduct with a smaller error and higher accuracy for URTE and IRTE through independent prediction. Thus it can provide a new means for medical meteorological forecast and is worth further research.