基于SAS的反向传播人工神经网络在国境卫生检疫领域中的应用研究

    裘炯良1，郑剑宁2，施惠祥1，张锜1

    1.宁波出入境检验检疫局卫生检疫监管处，浙江 宁波315012；2.北仑出入境检验检疫局

    摘要：目的 探索反向传播（BP）人工神经网络在国境卫生检疫领域中的应用研究。方法采用18×5×1结构的3层BP神经网络模型，对2007年到达宁波港国际航行船舶中的媒介阳性船舶170艘和对照船舶680艘进行数据训练和验证，并以建立的神经网络模型预测新到港的船舶外来媒介携带率。结果经过100次的迭代运算，训练过程的误判率为0.1647，验证过程的误判率为0.1824；训练过程的平均误差为0.3668，而验证过程的平均误差为0.4550。通过该神经网络模型预测船舶携带外来媒介情况与实际结果的符合率达到83.3%，预测效果良好。结论针对高度不确定的非线性系统，应用BP人工神经网络可实现相对精确的预测功能，为国境卫生检疫风险评估及预警方面的研究提供理论基础。

    关键词：神经网络；外来媒介；预测；反向传播

    中图分类号：R183.5 R184.1　文献标识码：B

    Application of Back Propagation NeuralNetwork

    on health quarantine based on SAS

    QIU Jiong-liang*, ZHENG Jian-ning, SHIHui-xiang, ZHANG Qi

    *Ningbo Entry-exit Inspection and QuarantineBureau, Zhejiang, Ningbo 315012, China

    Abstract:   Objective  To explore the application of Back Propagation Neural Network onhealth quarantine based on SAS.   Methods   TheBack Propagation Neural Network (BPNN)with the structure of 18×5×1was employed for the calculation. A total of 170 vessels withpossive exortic vectors and 680 ones with negative vectors were putinto the BP neural network. And the messages about new arrivalvessels were used for the prediction by the network.  Results   After one hundred time of iteration,misclassification rate of the training was 0.1647 with the 0.3668average error; while misclassification rate of the validation was0.1824 with the 0.4550 average error. The predictive condition wasgood as the according rate attained 83.3%.  Conclusion   We can execute the relatively exactprediction based on BP neural network, especially for the highlyuncertain nonlinear system. So the network can provide thetheoretical base for the risk analysis and alert of healthquarantine.

    Key words:   Neural Network; Exoticmedical vectors; Prediction; Back Propagation

    《中国国境卫生检疫杂志》2012年第6期