基于相似性分析的SVM快速分类算法

Fast SVM Classification Algorithm Based on Similarity Analysis

针对支持向量机(SVM)分类速度取决于支持向量数目的应用瓶颈,提出一种SVM快速分类算法。通过引入支持向量在特征空间的相似性度量,构建特征空间中的最小支撑树,在此基础上将支持向量按相似性最大进行分组,依次在每组中找到决定因子和调整因子,用两者的线性组合拟合一组支持向量在特征空间的加权和,以减少支持向量的数量,提高支持向量机的分类速度。实验结果证明,该方法能以很小的分类精度损失换取较大的分类时间缩减,满足SVM实时分类的要求。

Aiming at the bottleneck of SVM that the speed of classification depends on the number of support vectors,this paper proposes a fast classification algorithm for SVM.In feature space it constructs the minimum spanning by introducing the similarity measure and divides the support vectors into groups according to the maximum similarity.The determinant factor and the adjusting factor are found in each group by some rules.In order to simplify the support vectors,it takes the linear combination of determinant factor and adjusting factor to fit the weighted sums of support vectors in feature space,so that the speed of classification is improved.Experimental results show that the algorithm can get higher reduction rate of classification time by minor loss of classification accuracy and it can satisfy the requirements of real-time classification.