Automated Data Pre-Processing
As indicated in Chapter 1 above, many learning methods require partition of continuous attributes (features) into discrete intervals. Such methods include neural networks, Bayesian models and standard decision tree algorithms (e.g., 1D3 and C4.5). Since some attributes in real-world databases may be continuous, several methods for discretizing these attributes have been developed.
KeywordsMutual Information Continuous Attribute Discretization Method Target Attribute Conditional Mutual Information
Unable to display preview. Download preview PDF.