A more efficient classification scheme for ID3

Abstract

Analyzed the principles and implementation steps of ID3 and the existing two improved ID3 algorithms proposed in [4] and [5]. ID3 has the shortcoming of inclining to choose attributes which have many values. Although the current two improved classification algorithms have solved the shortcoming of ID3, their classification time is not enough short, and their classification accuracy is not enough high. We proposed a new scheme. Our scheme solved the shortcoming of ID3 and improved the existing two algorithms effectively. Finally we use experiment to prove that the new scheme has shorter classification time and higher classification accuracy than ID3 and the existing two classification algorithms.