Decision trees, knowledge rules and some related data mining algorithms

Abstract

We show that the decision tree representation and the knowledge rules representation for data mining are semantically equivalent. Quinlan's production rule generators use attribute removal functions that are more powerful than ID3. The HCC-algorithm uses both attribute removal and concept tree ascension in its generalization function. The HCC-algorithm usually has more generalization power than the other ones. On computational efficiency, ID3 and the HCC-algorithm are efficient while Quinlan's production rule generators are less efficient. We also point out some disadvantages of the HCC-algorithm.

We propose a hybrid algorithm that combines all the above generalization functions and at the same time avoids over-generalization carefully. It is stable, complete, and non-local and is about as efficient as Quinlan's rule generators.