Steps of CART: Classification and Regression Tree

746 Words3 Pages
Overview The classic CART: Classification and Regression Tree algorithm was created by Breiman. The CART method is binary recursive partitioning procedure that can be used to process both continuous and nominal attributes as targets and predictors. The binary splits are the splitting of the data represented by nodes; each node is split into two child nodes to represent the binary split on the data into two separate paths. The recursive part of CART means that any child nodes can be additionally split into more children nodes and so forth. The partitioning refers to the data being split into multiple sections along the nodes into classifications. CART Four Main Steps The CART method involves four steps: tree creation, stopping the tree creation, tree pruning, and selecting the optimal tree. Tree creation begins with a root node, which includes all the data in the training set. CART then selects the best variable to split this node into two child nodes. To find the best variable to perform this split, CART runs through all variables and their values to determine the best combination to split the node on. The process of node splitting is repeated for each created child node and continued recursively until a stopping point is reached where no further splitting can be performed. Each node within this process is assigned a predicted class based on three factors, the assumed probability of each class, a cost matrix, and the percentage of the classifier that are located in each
Open Document