Attribute reduction algorithm based on new conditional information quantity

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 consistent and 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.