New method about how to construct decision tree based on association rule

Abstract

Nowadays, decision tree is widely used as one of the most powerful tools in data mining. However, to construct an optimization decision tree is a complete NP problem. So a new method about how to construct decision tree, which is based on association rule mining, is proposed in this paper. Firstly, approximate exact rule with high reliability is defined. Secondly new attributes are generated from the approximate exact rule. And then its evaluation method is discussed in detail. Thirdly, the decision tree is constructed with both the new generated attributes and its original data. Finally, after comprehensive analysis, experimental results show that this new method has higher accuracy than any other old method.