Simplify Multi-valued Decision Trees

Abstract

Decision tree is one of the popular data, mining algorithms and it has been applied oil many classification application areas. lit many applications, the number of attribute values may be over hundreds and that will be difficult, to analyze the result. The purpose Of this paper will focus on the construction of categorical decision trees. A binary splitting decision tree algorithm is proposed to simplify the classification outcomes. It adopts the complement operation to simplify the split of interior nodes and it is suitable to apply on the decision trees where the number of outcomes is numerous. lit addition, meta-attribute could be applied oil some applications where the number of outcomes is numerous and the meta-attribute is meaningful. The benefit of meta-attribute representation is that it could transfer the original attributes into higher level concepts and that could reduce the number of outcomes.