A new way to choose splitting attribute in ID3 algorithm

Aimed at solving the problem of tendency to multi-value attribute and huge computational complexity in ID3 algorithm, we proposed a new way to choose the splitting attribute. And a new conception of consistency is introduced in this paper based on rough set, we use it as certification of splitting data set. The decision tree is established according to each attribute's consistency rather than information entropy gain. In this way, it can avoid the problem of tendency to multi-value and huge computational complexity in traditional ID3 algorithm. We tested the improved algorithm, using three UCI Machine Learning Repository data sets, and compare the improved algorithm with the traditional ID3 algorithm. the experiment result shows that the accuracy rate of improved algorithm is higher than traditional ID3 algorithm.