SVM-KNN分类算法研究

SVM-KNN分类算法是一种将支持向量机（SVM）分类和最近邻（NN）分类相结合的新分类方法。针对传统SVM分类器中存在的问题,该算法通过支持向量机的序列最小优化（SMO）训练算法对数据集进行训练,将距离差小于给定阈值的样本代入以每类所有的支持向量作为代表点的K近邻分类器中进行分类。在UCI数据集上的实验结果表明,该分类器的分类准确率比单纯使用SVM分类器要高,它在一定程度上不受核函数参数选择的影响,具有较好的稳健性。

Study on SVM-KNN Classification Algorithm

SVM-KNN classification algorithm is a new combinative classification method,which is formed with support vector machine and nearest neighborhood.With the problem of the traditional SVM classifier,the training data can get a best classification hyperplane through sequential minimal optimization（SMO） algorithm,then use KNN classification algorithm with all support vectors in every class as representative points.The experimental results in the UCI data sets show that the classification accuracy rate of the new classifier has some increase than only using the SVM classifier.