基于SVM的粮食霉变预测分类方法研究

Research on Prediction and Classification Method of Grain Mildew Based on SVM

霉变是导致粮食储藏过程中数量减少、质量降低的重要因素,若能早期预测粮食是否会发生霉变,提前采取处置措施,对保障粮食储藏安全,降低粮食损失具有重要的意义。本研究采用支持向量机算法,并通过网格搜索优化参数,分别建立了稻谷和小麦霉变的预测分类模型,以判定在给定水分、温度和储藏时间的条件下是否会发生霉变。实验结果表示,稻谷平均准确率可达96%以上,小麦平均准确率可达92%以上。同时本研究采取不同规模的小样本训练建模,并与BP神经网络模型进行对比,训练结果表明,基于SVM的模型准确率高且表现稳定,明显优于BP神经网络模型。

Mildew is an important factor to reduce the quantity and quality of grain in the process of grain storage.It would be of great significance to ensure the safety of grain storage and reduce the loss of grain if we can early predict whether the grain will mildew or not,and take measures to deal with it in advance.In this paper,the prediction and classification models of paddy and wheat mildew respectively through support vector algorithm with Grid Search optimization parameter were established in this paper,in order to determine whether mildew would occur under the conditions of given moisture,temperature and storage time.The experimental results showed that the average accuracy of paddy was more than 96%,and that of wheat was more than 92%.At the same time,this study adopts different scale of small sample training modeling,and compared with BP neural network model,the training results showed that the SVM based model has high accuracy and stable performance,obviously better than the BP neural network model.