Abstract: Bayesian rough set(BRS) model for two decision classes is extended to more decision classes based on the analysis of BRS model. Corresponding definitions and simple property are given. At the same time, attribute significance is measured from the sides of global relative gain. Heuristic algorithm is proposed which regardes global relative gain as the heuristic information. Corresponding algorithm is run by the programming of Matlab. Both the attribute reduction of VPRS model and the attribute reduction of BRS model are compared by the results of the practical example. The validity and feasibility of the algorithm of BRS model are proved.

Key words: variable precision rough set model, Bayesian rough set model, R reduction, prior probability

摘要： 在分析贝叶斯粗糙集模型的基础上，将只含有两个决策类的贝叶斯粗糙集的情况推广至含有多个决策类的情况，给出了相关定义和简单性质。从全局相对增益的角度分析了属性重要度，给出以此为启发式信息贝叶斯粗糙集属性约简的启发式算法，且用相应的Matlab程序进行实现。对贝叶斯粗糙集与变精度粗糙属性约简进行了比较，结果证明了算法的有效性。

关键词: 变精度粗糙集模型, 贝叶斯粗糙集模型, R约简, 先验概率

CLC Number: 

• TP18