随着人民生活质量的改善，对农产品的质量安全意识越来越高，农产品的种类、安全因素和流通环节复杂，建立全过程的监管与追溯非常重要，而大数据技术的迅猛发展更是为农产品质量监管追溯系统提供了新的平台。 在分析国内外农产品质量监管方面和追溯系统技术的基础上，利用大数据技术，基于Hadoop平台架构了农产品质量监管与追溯系统。通过对SVM和BP神经网络算法的深入分析，构建了基于SVM算法的农产品区域质量监管预测模型，在选择最佳的惩罚因子和核函数参数时，将原始数据中的农药污染指数和重金属污染指数平均分为K组，每组数据分别做一次验证集，剩下的K-1组数据作为训练集，用验证集的分类精度的平均值作为分类器最终的交叉验证精度，用最大精度对应的惩罚因子和核函数参数进行训练，实现对需要重点监管的农产品区域的预测，与BP神经网络算法进行对比，SVM算法的分类准确率提高了10%；为了实现对农产品腐败率数据未来一段时间的预测，构建了基于SVR算法的农产品时序质量预测模型，将农产品近期的腐败率数据分为两组，最佳的惩罚因子和核函数参数选取与上一模型相同，用第一组数据训练得到的模型进行预测，预测数据与第二组真实数据进行绝对误差和相对误差分析，与BP神经网络算法进行对比，在预测值和真实值的相关系数方面SVR算法比BP神经网络算法提高了近5%，更加逼近真实值。 从Web端和Android端的角度对农产品质量监管与追溯系统进行设计。搭建测试环境，分别对Hadoop监管追溯平台、Web端和Android端的主要功能进行了测试，测试结果表明该系统在农产品质量监管与追溯方面具有一定的实用价值。

With the improvement of living quality, security consciousness of agricultural products quality becomes higher and higher. Due to the complicated types, safety factors and circulation of gricultural products, it is very important to establish the whole process of regulation and traceability. The rapid development of big data technology provides a new platform for the quality supervision and traceability system of agricultural products. On the basis of analyzing domestic and foreign quality supervision and traceability system technologies of agricultural products, an agricultural product quality supervision and traceability system is set up based on Hadoop platform with big data technology. Based on the deep analysis of support vector machine and BP neural network, the regional quality supervision and forecast model of agricultural products is established base on SVM. When choosing the penalty factor and kernel function parameter, the pesticide pollution index and heavy metal pollution index in the original data are divided into K groups. Each group of data is used as a validation set, and the remaining K-1 group data is used as the training set. The average classification precision of validation set is used as the final cross validation accuracy of the classifier. The penalty factor and kernel function parameter corresponding to the maximum precision is used to train the model, realizing the prediction of agricultural products region that needs to be supervised. Compared with BP neural network, the classification accuracy of SVM has improved by 10 percent. In order to realize the prediction of agricultural products corruption rate data, the timing quality prediction model of agricultural products was constructed based on SVR. The corruption rate of agricultural products were divided into two groups. The selection method of optimal penalty factor and kernel function parameter is same as the previous model. The model is trained with the first group of data and predict the data with the second original data, analyzing absolute error and relative error. Compared with BP neural network, the SVR algorithm improved nearly 5 percent on correlation coefficient between the predicted value and the true value and is closer to the original data. Agricultural products quality supervision and traceability system is designed from Web and Android. The test environment is built. The main functions of the Hadoop supervision traceability platform, Web and Android end are tested. The test results show that the system has certain practical value in agricultural product quality supervision and traceability.