An Efficient Knowledge Reduction Algorithm Based on New Conditional Information Entropy
Guowei Yang · Kongzhi yu juece · 2005
The disadvantages of the current conditional information entropy are analyzed. A new conditional information entropy is proposed. Based on this entropy the new significance of an attribute is defined and compared with two significances of this attribute based on the positive region and the current conditional information entropy respectively. The result shows that when used as heuristic information, the proposed significance of the attribute is better than the other two. Finally, a heuristic algorithm for knowledge reduction is designed and an efficient algorithm for computing conditional information entropy is proposed. Theoretical analysis and experimental results show that time complexity of this reduction algorithm is less than that of the algorithm based on the current conditional information entropy. Also, this reduction algorithm is more capable of finding the minimal or optimal reducts.