An improved decision tree algorithm based on mutual information

As a classical data mining algorithm, decision tree has a wide range of application areas. Most of the researches on decision tree are based on ID3 and its derivative algorithms, which are all based on information entropy. In this paper, as the most important key point of the decision tree, the metric of the split attribute is studied. The mutual information is introduced into decision tree classification. The results show that the decision tree classification model based on mutual information is a better classifier. Compared with the ID3 classifier based on information entropy, it is verified that the accuracy of the decision tree algorithm based on mutual information has been greatly improved, and the construction of the classifier is more rapid.