一种基于粗集理论的增量式属性约简算法

增量式学习中,当信息系统的对象和决策属性不变而不断增加条件属性时,为了获得该系统的约简属性,一般方法是对决策表中的所有数据重新计算,但这种方法显然效率很低且不必要.在粗集理论的基础上,给出相对区分矩阵和绝对区分矩阵的定义,提出一种新的增量式属性约简算法.通过实例得知:由该算法得到的属性约简与传统算法得到的属性约简结果相同,但该算法不仅降低了时间复杂度而且其分类质量一般要优于原来的分类质量,所以该属性约简具有一定的实用价值.

AN INCREMENTAL ALGORITHM FOR ATTRIBUTE REDUCTION BASED ON ROUGH SETS THEORY

In order to obtain the reduced attributes of an information system when some condition attributes are added constantly to the information system without the change of object and decision attribute for incremental learning,all the data in the decision table will be recalculated in the classical method.Clearly,this method is not effective and necessary.In this paper,the definitions of relative distinction matrix and absolute distinction matrix are given.A new incremental algorithm for attribute reduction base...