Abstract: This paper uses the information entropy method to measure the rough set’s fuzziness，and makes the pretreatment before the reduction of rough’s decision attribute with eliminating the difference which due to the redundance of decision attribute.Combination of SVM in solving the small sample，nonlinear and high dimensional pattern recognition problem has a lot of unique performance advantages.In this paper，an improved algorithm is given and the classification results are tested.

Key words: information entropy, rough set, fuzzy degree, reduction, Support Vector Machine（SVM）

摘要： 采用信息熵的方法来度量粗糙集的模糊性可以在约简之前对粗糙的决策属性进行预处理，从而消除因决策属性的冗余而带来的分类决策的偏差。结合 SVM在解决小样本、非线性及高维模式识别问题中表现出许多特有的优势。对该类别的方法进行了改进，并对分类的结果进行了测试。

关键词: 信息熵, 粗糙集, 模糊度, 约简, 支持向量机