为提高支持向量机集成的泛化性能,提出一种基于独立成分分析法的特征Bagging支持向量机集成方法,删除了冗余特征.该方法从得到的独立成分特征空间中提取特征子空间,避免了直接从原特征空间中随机选择特征子空间而导致的对特征依赖或相关性的破坏,提高了个体支持向量机的性能,保证了个体支持向量机之间的差异度.在UCI和Stat-Log数据集合上的仿真实验表明,该方法具有更好的泛化性能. In order to improve the generalization performance of support vector machine (SVM), an integrated method of feature Bagging SVM based on independent component analysis is proposed to remove the redundant features. The method extracts the feature subspace from the independent component feature space and avoids In order to improve the performance of individual SVM and to ensure the difference between individual SVMs, the disruption of feature dependence or correlation caused by the random selection of feature subspace directly from the original feature space, Simulation experiments on Log data set show that this method has better generalization performance.