Abstract: According to the fact that the bootstrap in ensemble learning can’t generate the committee classifiers with big differences，automatical weighted SVM ensemble learning based on multimodal perturbation is proposed（MP-AWE）.MP-AWE generates the perturbation on the training data with bootstrap sampling，generates the perturbation on the input attributes with PCA attribute subspace selection，and generates the perturbation on the learning parameter with automatic model selection，and generates the perturbation on the output with accuracy of the committee classifiers.The experimental results show that the performance of MP-AWE is better than that of many other ensemble algorithms.

Key words: ensemble learning, Support Vector Machine（SVM）, automatic model selection, multimodal perturbation, Principal Component Analysis（PCA）

摘要： 针对集成学习中bootstrap方法不能产生具有较大差异性的成员分类器，提出基于多模式扰动模型动态加权SVM集成方法。该方法在训练样本中使用bootstrap采样产生扰动，在输入特征中使用PCA特征滤波子空间法产生扰动，用自动模型选择法来动态扰动每个成员分类器的参数，用分类精度对成员分类器加权集成扰动输出。实验结果表明该方法比常用的bootstrap集成方法具有更好的集成效果。

关键词: 集成学习, 支持向量机, 自动模型选择, 多模式扰动, 主成分分析