一种基于条件熵的决策表属性约简算法

AN ALGORITHM OF ATTRIBUTES REDUCTION OF DECISION TABLE BASED ON CONDITIONAL ENTROPY

以分类为基础提出一种基于条件熵的决策表属性约简算法。通过条件熵的计算在属性约简的同时将原决策表逐层分解成相对于决策属性来说尽量均匀的子决策表,从而缩小了数据规模;随后对算法的时间复杂度进行了分析;实验表明,该算法在效率方面优于传统算法。

This paper proposes an algorithm of attributes reduction of decision table based on conditional entropy on the basis of classification.Through the calculation of the conditional entropy,this algorithm reduces the attributes while disassembles the decision table tier by tier into sub-tables as symmetrical relatively to the decision attributes as possible,therefore the size of data is dwindled.The time complexity of the algorithm is also analyzed afterwards in the paper.Experiment shows that the new algorithm outperforms the traditional algorithm in efficiency.