Decision trees are a common procedure used in data mining and supervised learning because of their robustness to many of the problems of real world data, such as missing values, irrelevant variables, outliers in input variables, and variable scalings. The algorithm is an “off-the-shelf” procedure, with a few parameters to tune.
This implementation creates only one decision tree, as opposed to generating multiple trees, as in the case of random forests. These functions support classification trees on continuous variables and categorical attributes, handle missing values during the prediction phase, and support GINI, entropy, and chi-square impurity measurements.