用属性单值表示的决策表简化算法及属性核计算

Decision Table Simplification Algorithm Based on Attribute Single-Valued Representation and Attribute Core Calculation

为了降低决策表的存储空间,提高计算属性核的效率,提出了一种用属性单值表示的决策表简化算法.给出了条件属性的尺度、权值、属性单值和简化决策表的定义,以及尺度运算、权值运算、每个对象的条件属性值集合转换为一个属性单值的简化运算的严密公式,并通过决策表的可逆定理证明了简化决策表与原始决策表具有相同的信息表达能力.利用该算法简化决策表,条件属性的表达由多维降低为一维,从而有效地降低了决策表的存储空间.实验结果表明,简化决策表属性核的计算效率明显比现有决策表属性核的计算效率高,尤其是对于大型决策表,这种优势更加明显.

To reduce the decision table of storage space,and improve the efficiency of computing attribute core,a decision table simplification algorithm based on attribute single-valued representation is proposed.The definitions of the size and weight of condition attribute,single-valued attribute and simplified decision table are presented.Then the rigorous calculation formulas of the size,weight and the condition attribute value set of every object converting into a single-value are given.The inverse theorem is proved,which indicates that the simplified decision table are endowed the same information with the original decision table.The storage space of decision table is reduced greatly because the multi-dimension attribute values are decreased to one dimension ones.Then the attribute core algorithm based on the simplified decision table is compared with that based on existing decision table.Experimental results show that the computing efficiency of the former is obviously higher than that of the latter,especially for large-scale decision table.