基于云模型的连续属性决策表简化算法

Reduction algorithm for decision table with continuous attributes based on cloud model

从降低样本数据模糊性及随机性角度出发,提出了一种基于云模型的连续属性决策表简化算法.该算法通过对决策表的转换,建立云相似度概念来刻画样本间等价关系,并采用改进的动态聚类方法自动获取相似样本,完成对样本的较粗粒度表示.实验结果表明,简化决策表在不到原规模十分之一的条件下,取得了大致相当的识别率,从而大幅降低粗集知识获取时间.

Decision table is a kind of special and important knowledge system which has been applied in the decision support and data mining fields widely. From the point of reducing the fuzziness and randomness of samples, a reduction algorithm for decision table with continuous attributes based on cloud model is presented. By transformation in decision table, it constructs cloud similarity to describe the equivalence relation among samples, and obtains automatically the similar samples by using the improved dynamic clustering method. Finally, the similar samples are expressed as a thicker granularity. The experimental results show that similar recognition rate has been obtained in less than one-tenth of the original scale of decision table. On this basis, the knowledge acquisition time using rough set is substantially reduced.