支持向量机方法在烟叶可用性预测中的应用

摘要

In order to evaluate the classification of tobacco leaf usability, the prediction model of the classification in tobacco leaf usability was established by using support vector machine (SVM) method with different kernel functions, and compare accuracy of the predicted result with Fisher methods. The results indicated the prediction accuracy with SVM method was higher than that of Fisher method. Among other kernel functions based on RBF, The SVM model was the best, and the prediction accuracy reached 90%. Therefore, the SVM method was an effective tool for classifying and predicting tobacco leaf usability.%为了对烟叶可用性分类进行评价,采用不同的核函数建立烟叶可用性支持向量机(Support Vector Machine,SVM)预测模型,对预测集样本进行预测,并与Fisher法的预测结果进行了比较.结果表明,SVM算法所建立的数学模型的预测准确率均比Fisher法高,且以径向基函数(RBF)建立的SVM分类模型的预测效果最好,对预测集样本的准确率达90％,说明SVM分类模型能较好地预测烟叶可用性.