基于条件信息量的快速粗集约简算法

Rapid Reduction Algorithm Based on the Conditional Information Quantity

为提高粗集约简的效率,提出了一种基于条件信息量的快速粗集约简算法.该算法定义了基于等价块的信息量与条件信息量,并给出了条件信息量的不变性定理与属性重要度的单调性定理.与其他算法相比较,该算法缩小了实例集合的规模,减少了需要计算重要度的属性个数.实验结果表明,保持约简集合不变的前提下,该算法有效提高了粗集约简的效率.

To improve the efficiency of attribute reduction, a rapid reduction algorithm based on conditional information quantity is proposed. The concepts of information quantity based on partition and conditional information quantity based on partition are defined, and the theorems about the monotone of attribute significance^invariance of conditional information quantity based on partition are proven; Compared with other algorithms, this algorithm reduces the searching space of attributes and samples in each step. Experimental results showed that the rapid reduction algorithm is more efficient than the existing algorithms.