基于长短期记忆神经网络的重大活动中鲜(冻)肉制品铅含量风险预测

目的 针对未来时间内重大活动举办地鲜(冻)肉制品铅含量风险预测的问题,建立基于LSTM的时间序列预测模型,对当地鲜(冻)肉制品铅含量进行风险评估与预测预警。方法 通过收集2011-2020年国家市场监督管理总局日常食品监督管理抽检数据,筛选出北京的鲜(冻)肉制品的抽检数据,构建数据集并进行预处理,按照7:2:1的比例划分训练集、验证集和测试集,基于Tensorflow平台构建4层LSTM模型并进行训练,基于重大活动举办前10天的鲜(冻)制品铅含量数据,对未来1天的鲜(冻)制品铅含量风险进行预测。结果 实验表明,经过50轮模型迭代训练,训练集和测试集Loss指标收敛至0.084,经过5次训练后的模型评估参数RMSE为0.192,R2_score为0.916,模型误差较小、准确度较高。结论 基于LSTM的鲜(冻)肉制品铅含量风险预测模型整体性能较好,可应用于重大活动举办地的食品风险预测,并精准指导监督抽检。

Risk prediction of lead content in fresh (frozen) meat products in major activities based on long short-term memory neural network

Objective To establish a time series forecasting model based on long short-term memory (LSTM) neural network, and carry out risk assessment, prediction and early warning of lead content in fresh (frozen) meat products in Beijing. Methods Through collecting the sampling data of daily food supervision and management of the State General Administration of Market Supervision and Administration from 2011 to 2020, the sampling data of fresh (frozen) meat products in Beijing were screened out, and the data set was constructed and pretreated, the training set, verification set and test set were divided according to the ratio of 7:2:1, a 4-layer LSTM model was constructed and trained based on Tensorflow platform, and the lead content risk of fresh (frozen) meat products in the next 1 d was predicted based on the lead content data of fresh (frozen) meat products in the 10 d before the major events. Results After 50 rounds of model iterative training, the Loss index of the training set and the test set converged to 0.084, the root mean square error (RMSE) of the model evaluation parameters after 5 times of training was 0.192, and the coefficient of determination (R2_score) was 0.916, and the model had small error and high accuracy. Conclusion The LSTM based lead content risk prediction model of fresh (frozen) meat products has good overall performance, and can be applied to the food risk prediction of major events, and accurately guide the supervision and sampling inspection.