一个混合特征属性选择算法

A Mixing Algorithm for Feature Attribute Selection

1.引言特征属性选择(feature attribute selection,FAS)是机器学习和模式识别中比较困难而又非常有意义的一个问题。FAS问题是从一个大的侯选属性集合中选择一个较好的、有代表性的属性子集。由于在实际应用中,过多的属性会严重影响归纳学习的质量,一些不必要的属性会加大训练数据量,影响学习速度,损害所生成规则的精度,因此FAS是一个有实际意义的问题。

The feature attribute selection is a very interesting problem. With the development of Rough Set theory(RS)during these years ,many researchers and scholars proposed the attribute selection based on RS. But with the increasement of the attribute number,the efficiency declines rapidly. In this paper, we combine the RS theory with GA and propose a mixing heuristic algorithm for attribute selection. The experiment result shows that it can get better result and higher efficiency especially for settling the problem of large attribute number.