A classification tree is used to predict qualitative variables instead of quantitative variables. It is very similar to a regression tree and uses recursive binary splitting to generate the tree. However, it uses the Gini index ($G = \sum_{k=1}^{K} \hat{p}_{mk} (1-\hat{p}_{mk})$) or entropy ($D = -\sum_{k=1}^{K} \hat{p}_{mk} \log \hat{p}_{mk}$) to evaluate splits. A small value indicates high node purity, meaning the node contains observations from almost the same category. If the $m$th node is pure, its entropy and Gini index are very small.