一种区间属性值离散化的新方法

A New Discretization Algorithm for Continuous Interval-valued Attributes

粗糙集理论作为一种新的处理含糊和不确定性问题的数学工具,已成为国际学术界的一个前沿的研究领域。传统的粗糙集理论只能对数据库中的离散属性进行处理,因此,连续属性值的离散化问题不容忽视。已有的离散化方法主要是针对固定点上的连续属性值的,实际应用中大量存在着连续区间属性值的情况。文中针对这一问题,提出了一种连续区间属性值离散化的新方法,并利用辐射源信号进行了仿真试验。结果表明,该方法能有效离散区间属性,从而拓展了粗糙集理论的应用范围。

Rough set theory plays an important role in knowledge discovery, and is regarded as a field of leading edge. But it cannot deal with continuous attributes, thus discretization is a problem which we cannot neglect. Consistency must be satisfied for discretization of decision systems in rough set theory. Existing discretization methods cannot well process continuous interval-valued attributes. A new approach is proposed to discretize continuous interval-valued attributes in this paper, which enhances accurate recognition rate in pattern recognition. At last, examples of recognizing the emitter purpose are selected to demonstrate this new method. Experimental results show that the performance of this new method is accurate and effective. So the approach expands the application scope of rough set theory.