一种基于约简树的决策表属性约简方法

本文首先给出了约简树的定义，提出了采用宽度优先的策略构造约简树，在构造约简树的过程中。把某个结点处决策规则是杏一致作为在该结点处是否可以继续构造子树的条件，把该结点是否已经出现过作为剪枝条件，并把相对核作为构违约简树的一个重要的终止条件。随后研究了约简树的性质，找到了基于约简树的决策表的极小属性集。最后分析了该方法在最好和最坏情况下的时间复杂度。实例分析表明，该算法能有效地对决策表进行属性约简。

An Attribute Reduction Approach for Decision Table Based on Reduction Tree

The definition of reduction tree and an approach on building the reduction tree applying the breadth -first strategy were given. Whether the decision rules were consistent in one node was regarded as the condition of building a subtree whose root node was this node. Whether one node existed was regarded as the condition of pruning the branch in the node and the relative core was regard as the important condition of ending the reduction tree. And then the characters of the reduction tree were researched and the minimal attribute reduction was found in the reduction tree. Finally, the time complexity in the best and the worst case was analyzed. The experimental results showed that this algorithm was effective in attribute reduction of decision tables.