ChiMerge: Discretization of Numeric Attributes

Learning

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 10

Learning: Inductive

Many classification algorithms require that the training data contain only discrete attributes. To use such an algorithm when there are numeric attributes, all numeric values must first be converted into discrete values-a process called discretization. This paper describes ChiMerge, a general, robust algorithm that uses the x2 statistic to discretize (quantize) numeric attributes.