一种具有双控制比例因子的最优间隔超球分类器

An optimal separation hyper-sphere classification model with double proportion control parameters

针对支持向量描述只考虑目标类训练样本,结合支持向量机最优分类超平面和支持向量描述的思想,引入了异常样本信息的监督机制,建立了最优间隔超球分类器模型,以一个最小的超球包含目标类训练样本和一个尽可能大的超球体将非目标样本隔离在超球体外,使决策超球面与该两个超球面以最大间隔分离,保证了描述精度和泛化性能,同时,为更好地排除对两类样本数据分布中野点的干扰,提出了一种双控制比例因子的控制方法,更加灵活地实现软间隔分类,仿真实例验证了该分类器具有比SVDD更好的分类性能。

After analyzing the disadvantage of unsupervised training of support vector data description （SVDD）, combining the advantage of optimal separation hyper-plane and SVDD, and inducing the supervision of information of negative class, a hyper-sphere classification model with optimal separation was proposed. With one minimum hyper-sphere containing positive class and one hyper-sphere as big as possible excluding negative class, the decision hyper-sphere was made to separate itself and the two hyper-spheres with the max distance to improve the model＇s description accuracy and generalization performance. To remove the interference of bad points, a method with double proportion control parameter was proposed, it could realize soft separation. Simulation results of Banana and UCI data sets showed that the proposed model has better classification performance than SVDD.