基于监测数据和BP神经网络的食品安全预警模型

以中国实际食品安全监测数据为样本，研究基于BP神经网络的食品安全预警方法。首先对食品安全日常监测数据进行筛选简化，选择其中与食品安全最为密切的167种检测项目，以此检测项目为指标按月度划分建立数据样本。然后建立以167种检测项为输入层，包含2个隐层，以化学污染、农药残留、兽药残留、重金属、微生物致病菌5大类为输出层的食品安全预警神经网络模型，最后用所得数据样本进行训练和验证。结果表明，基于BP神经网络的食品安全预警方法能有效识别、记忆食品危险特征，能够对输入样本进行有效的预测，研究有助于丰富食品安全数据的处理方法，有助于完善相关预警技术手段。

Model for food safety warning based on inspection data and BP neural network

Based on BP neural network theory, the food safety research was carried out with the daily food inspection data from Chinese General Administration of Quality Supervision,Inspection and Quarantine. Firstly, the inspection data was simplified to 167 supervised items which had the most direct relation with food safety forecast. Then the BP neural network model was established with input layer of the previous 167 items, five groups as output layer, and two hidden layers as passing function. Last, the model was trained and validated by the simplified dataset. The research showed that the model could effectively remember and identify characteristics of food inspection datasets and then make effective forecast for new dataset. This will be beneficial for research methodologies and techniques in Chinese food safety warning practice.