基于K最近邻决策的支持向量机分类算法及仿真

Algorithm and Simulation of SVM Classifier Based on KNN Judgment

将基于支持向量机(SVM)的分类方法和最近邻法(NN)相结合,提出了一种SVM-KNN的分类方法。通过SVM算法对训练样本进行训练并找出支持向量,在进行待识别样本判断时,当其与最优分类面距离大于某一给定阈值时采用SVM决策模型,否则运用K最近邻法决策其类别,从而减少SVM算法的误判概率。仿真实验结果显示,运用该算法无论对于合成数据还是真实数据,在分类精度上比单独的SVM都有较明显的提高。

A new algorithm of SVM-KNN is presented which combined the Support Vector Machine （SVM） with K Nearest Neighbor （KNN）. Firstly the training model is constituted and the support machine is found by SVM method through training sample. During the predicting phase, when the distance from the test sample to the optimal hyper plane is greater than the given threshold, the SVM model is applied to classified the test sample; otherwise the KNN would be used for reduce the misclassification probability. Simulation experimental results show that the SVM-KNN algorithm performs better than sole SVM in accuracy of classification through both the synthesized data and the real data.