【摘要】 针对单类数据的分类问题,提出一种基于支持向量数据描述(SVDD)的分类算法。该算法利用SVDD获得包含单类数据的最小球形边界,通过该边界对未知样本数据进行分类,同时采用可行方向方法求解边界优化中的二次规划问题,并在UCI机器学习数据集上将该算法与LS-SVM算法进行比较。实验结果表明,该算法不仅获得了更高的分类准确率,而且具有较低的运行时间。更多还原

【Abstract】 Aiming at the classification problem of one-class data, a classification algorithm based on Support Vector Data Description(SVDD) is proposed, which uses SVDD to obtain the minium sphericity boundary containing one-class data. The unknown sample data is classified with this boundary. In addition, the quadratic programming problem in boundary optimization is solved by using feasible direction method. This algorithm is also compared with LS-SVM in UCI dataset. Experimental results show this algorithm not only achieves higher classification accuracy, but also has lower running time.更多还原