论文部分内容阅读
粒计算是模拟人类思维和解决复杂问题的方法,它是复杂问题求解、海量数据挖掘、模糊信息处理的有效工具.文中首先分析并指出传统的规则获取方法存在的某些弊端,并从粒计算的角度分析属性约简的粒度原理,指出属性约简过程的本质是寻找决策划分空间的一个极大近似划分空间,而在极大近似划分空间上提取的规则可能不是最简规则.为此,提出一种基于最大粒的规则获取算法,该算法根据条件属性对论域形成的分层递阶的划分空间,自顶向下逐渐提取最大粒对应的规则.仿真实验表明该算法提高粗糙集的泛化能力.
Particle counting is a method to simulate human thinking and solve complex problems, and it is an effective tool to solve complex problems, mass data mining and fuzzy information processing.Firstly, the paper analyzes and points out some shortcomings of the traditional rules acquisition methods, The paper analyzes the granularity principle of attribute reduction and points out that the essence of attribute reduction process is to find a maximal approximate partition space of decision partition space and the rule extracted in maximal approximate partition space may not be the simplest rule.Therefore, This paper proposes a rule-based algorithm based on maximum grain, which extracts the rules corresponding to the maximum grain from the top to the bottom according to the hierarchical classification space formed by the conditional attributes on the universe of discourse. Simulation results show that this algorithm improves the rough set Generalization.