论文部分内容阅读
基于Rough Set理论中的不可分辨性原理,给出两个新的定义属性的最大区分值(Maximum Dis-cernibility Value,MDV)和属性冗余度(Attribute Redundancy Rate,ARR)。在数据预处理阶段,属性的MDV数值用于确定关于自组织映射网络SOM输出单元数量的启发式搜索策略;属性冗余度则用于衡量属性约简结果的信息冗余程度,并以此作为优化SOM网络输出层结构的依据。不依赖于领域经验知识,建立了MDV、SOM、ARR的组合算法模型,实现了Rough Set理论中连续属性的自动离散化计算,并明显提高了属性约简的速度。最后,通过项目实例对全过程进行有效验证。
Based on the principle of indivisibility in Rough Set theory, two new definition attributes, Maximum Discernibility Value (MDV) and Attribute Redundancy Rate (ARR), are given. In the data preprocessing stage, the attribute’s MDV value is used to determine the heuristic search strategy for the number of SOM output cells in the self-organizing map network. The attribute redundancy is used to measure the degree of information redundancy of the attribute reduction result and use it as The basis of optimizing SOM network output layer structure. Independent of domain knowledge, this paper establishes a combined algorithm model of MDV, SOM and ARR, realizes the automatic discretization of continuous attributes in Rough Set theory and obviously improves the speed of attribute reduction. Finally, the entire process is validated through project examples.