一种基于新的条件信息量的属性约简算法

Attribute reduction algorithm based on new conditional information quantity

为了获得决策系统中更好的相对约简,讨论了属性约简与条件信息量的关系,提出了新的条件信息量,由此定义新的属性重要性。统一了一致决策表和不一致决策表属性约简方法,以新的属性重要性为启发信息,给出了计算新的条件信息量的高效算法。理论分析和实验结果表明,与现有的基于条件信息量的约简算法相比,该算法时间复杂度较低,同时约简后的属性数目更少。

To obtain satisfactory relative attribute reduction in decision systems, the relationship between attribute reduction and conditional information quantity is discussed. A new conditional information quantity is proposed. Based on this quantity the new significance of an attribute is defined. The heuristic algorithm for attribute reduction is unified for consistentand inconsistent decision tables and an efficient algorithm for computing conditional information quantity is designed. Theoretical analysis and experimental results show that time complexity of this reduction algorithm is less than that of the existing algorithm based on the conditional information quantity, and that the number of attributes is small after the reduction.