K-means聚类与SVDD结合的新的分类算法

New classification algorithm K-means clustering combined with SVDD

为了提高支持向量数据描述(SVDD)的分类精度,引入局部疏密度提出了改进的SVDD算法。该算法提高了分类精度,但增加了计算复杂度。为此,先用K-means聚类将整个数据集划分为k个簇,再用改进的SVDD算法并行训练k个簇,最后再对获得的k个局部支持向量集训练,即得到最终的全局决策边界。由于采用了分而治之并行计算的方法,提高了算法的效率。对合成数据(200个)和实际数据的实验结果表明,所提算法较SVDD算法,训练时间降低为原来的10%,分类错误率较原来的降低了近一半。因此,所提算法提高了分类精度和算法效率。

This paper proposed an improved SVDD algorithm by introducing a local density degree for each data point in order to improve the support vector data description（SVDD） classification accuracy. Proved to improve the classification accuracy, but the increase of computational complexity. To this end, first divided the whole data set into k clusters using K-means cluste- ring algorithm. Then, trained the k clusters in parallel by improved SVDD. Finally, trained the k obtained local support vector sets and got the final overall decision border. As a result of divide and conquer method and parallel computing, improved the efficiency of the algorithm. Synthetic data and real data experimental results show that the proposed method than SVDD algorithm, training time is reduced to 10% and classification error rate lower than the original by almost half. Therefore, the proposed algorithm improves the classification accuracy and algorithm efficiency.