Decision Tree Induction for Dynamic, High-Dimensional Data Using P-Trees

1 Decision Tree Induction is a powerful classification tool that is much used in practice and works well for static data with dozens of attributes. We adapt the decision tree concept to a setting where data changes rapidly and hundreds or thousands of attributes may be relevant. Decision tree branches are evaluated as needed, based on the most recent data, focusing entirely on the data that needs to be classified. Our algorithm is based on the P-tree data structure that allows fast evaluation of counts of data points, and results in scaling that is better than linear in the data set size.

• Type: Conference
• Year: 2004
• Where: CATA
• Authors: Anne Denton, William Perrizo
• Keywords: CATA 2004, CATA 2007, Decision Tree, Decision Tree Concept, Decision Tree Induction