基于机器学习的微量农药光谱预测模型

A spectral prediction model of trace pesticide based on machine learning

以微量农药近红外光谱数据为研究对象,分别采用k近邻分类算法(k-nearest neighbor,k-NN)、朴素贝叶斯分类器(naive Bayesian classifier)、支持向量机(support vector machine,SVM)算法、极限学习机(extreme learning machine,ELM)等不同机器学习算法对其进行不同浓度分类判别分析。研究结果表明,k近邻分类算法、支持向量机算法、极限学习机算法均取得了较好的分类预测精度,分类预测精度均达到90%以上,其中极限学习机算法训练速度最快,对于大样本数据具有较好的解析精度和分析速度。机器学习算法为实现光谱快速分析检测提供了新的思路和有效解决办法。

The classification accuracy and speed of four machine learning techniques,including k-Nearest Neighbor(k-NN),Naive Bayes classifier,Support Vector Machine(SVM),and Extreme Learning Machine(ELM)are compared based on trace pesticide near-infrared spectroscopy.The results indicate that the classification accuracy of k-NN,SVM and ELM are better than that of Naive Bayes classifier,and the accuracy rate reaches more than 90%.Moreover,ELM performed much faster than k-NN and SVM,which indicates that ELM might be a promising method for real-time classification with a comparable accuracy based on near-infrared spectroscopy.