一种基于聚类的快速局部支持向量机算法

为进一步改善局部支持向量机的分类效率和分类精度,提出一种改进的局部支持向量机算法。该算法对每类训练样本分别进行聚类,使用聚类生成的样本中心点集代替样本,使用改进的k最近邻算法选取测试样本的k个近邻。分别在UCI数据集和自建树皮图像数据集上对本研究算法的有效性进行测试。实验结果表明,本研究提出的算法在分类精度和效率上具有一定的优势。

An algorithm of fast local support vector machine based on clustering

In order to further improve the classification efficiency and precision of local support vector machine, a new algorithm was proposed.The two major improvements were as follows. First, every type of training samples was clustered seperately, and the training samples were substituted for sample centers generated by clustering. Second, the k nearest neighbors of test samples were selected by using the improved k-nearest neighbor algorithm. Tests were done on UCI data sets and bark image data sets made by the proposed algorithm to verify its effectiveness. Experimental results demonstrated that this algorithm had certain superiority of classification accuracy and efficiency.