一种新的粗糙集与支持向量分类机算法

用支持向量机的机器学习是依据结构风险最小化原则，序列最小优化(SMO)是较特殊的分解算法。对高维大样本对象，支持向量机训练算法面临耗时增大与维数灾问题，利用粗糙集(RS)对不确定数据处理能力，提出一种新的粗糙集与支持向量分类机算法RS-SMO，可以对数据集做属性约简，生成类边界集作为SMO的训练子集，比原始训练集的维数与规模大小都有一定程度的减少，可构造出具有较好时空性能的算法。用两个实用数据对象做仿真，实验结果表明算法RS-SMO比SMO的性能有大的提高，实现了结构风险最小化。

A Novel Approach of Support Vector Classifier with Rough Set

Support Vector Machine obeys Structural Risk Minimization for machine learning. Sequence minimizing optimization (SMO) is an uttermost decomposing algorithm of SVM training. When training the object of high dimensions and large samples, SVM encounters dimension-tragedy and the problem of time-cost increase. As Rough Set has the capability of imprecise data management, a novel approach of Support Vector Classifier with Rough Set (RS-SMO) is presented. The database undergoes attribute-reduct. The boundary sets are formed as training-subset of SMO and are less dimensions and lengths than original training-set. The algorithm analyzing shows good performance of time and space. Two practice objects with huge-data are used to simulate. The experiment results validate RS-SMO outperforming SMO, and implementing Structural Risk Minimization.