An entropy-based discretization method for classification rules with inconsistency checking
Renpu Li, Wang Zheng-ou · 2003
Discretization is an effective technique in handling continuous attributes for data mining, especially for classification problems. Most entropy-based discretization methods are local and it is easy to lose valuable information in the data. We present an entropy-based algorithm. Through inconsistency checking, we may add/delete cut points on the basis of a preliminary discretization scheme. So the interaction between all attributes is taken into consideration in the discretization process which makes our method possess a global property. Experimental results indicate that with the same rule generator C4.5, our method can produce stronger rules than existing entropy-based discretization methods.