基于支持向量预选取的支持向量域故障预报

Support vector domain fault prediction based on support vector pre-extracting

针对两种支持向量域模型分别分析了支持向量的分布特性,在此基础上从训练集中选取具有一定几何特征的向量构建预测模型.这些特征向量的预选取在不影响支持向量域的故障预报能力的前提下,大大减少了训练样本,提高了支持向量域的训练效率.仿真实验表明了该方法的有效性和可行性.

Aiming at the two kinds of support vector domian models, the support vector＇s distribution characteristics are analyzed. On this basis, the prediction models are constructed by the vectors with some special geometrical character extracted from the training set. The pre-extracting method of the suppor vectors greatly reduces the training samples and speeds up the training speed of support vector domain, at the same time, the fault prediction ability is well maintained. The simulation results show the effectiveness and feasibility of the method.