Construction of decision trees based entropy and rough sets under tolerance relation

Decision tree induction is one of the most popular data mining techniques with applications in various fields. Present algorithms for construction decision trees can not deal with missing value in information systems properly. A new concept, rough gain ratio, is first introduced by the aid of tolerance relations in the extended rough sets theory. Then, an approach for inducing decision trees under the rough gain ratio is presented. Examples show that the decision trees generated by the proposed method tend to have simpler structure and more understandable rules than C4.5.