A classification tree predicts categorical outcomes. When presented with new data, it directs the data to a specific leaf node based on its characteristics. The prediction for that node is determined by the most frequent class label among the training samples that arrived at that node. This is because classification aims to identify the most probable class. Mean and median are utilized in regression trees for continuous outputs, not for classification. RMSE, a measure of performance for regression, is not a prediction method. Consequently, the mode is the appropriate method for classification trees.