In order to filter out irrelevant or lower degree of attributes, it must use attribute reduction algorithms, therefore, attribute reduction is a core research topic in rough sets . In attribute reduction algorithms based on discernible matrix, discernible matrix must be acquired firstly. But the space of storing the discernible matrix in computer is very difficulty when the scale of the problem is very large. Moreover, the computing cost of the algorithm is higher. In order to solve above-mentioned problems, the definition of discernible sets is firstly provided, and then a new attribute reduction algorithm based on the discernible sets is designed. The new attribute reduction algorithm need not create the discernible matrix and not make those unused elements in the attribution reduction progress, so it can cut down the computing and storing capacity greatly. The algorithm efficiency is illustrated with an experimental example.%为了过滤掉不相关或相关程度较低的属性,就必须使用属性约简算法,从而使得属性约简成为粗糙集中一个核心的研究课题.基于差别矩阵的属性约简算法求解时总是先要求出差别矩阵,当问题规模增大时,将导致存放差别矩阵的空间过大和算法执行时间过长.针对这一问题,本文提出辨识集的定义,并利用辨识集设计一个新的属性约简算法.新算法在属性约简过程中不生成差别矩阵和大量的无用元素,大大减少存储量和计算量,从而提高算法的效率.实验验证了新算法的高效性.
展开▼
