Abstract
Binning of continuous variables and cardinality reduction of nominal variables are two common transformations used to achieve two objectives—(1) reduce the complexity of independent and possibly dependent variables; and (2) improve the predictive power of the variable. By careful binning and grouping of categories, the predictive power of a variable with respect to a dependent variable for both estimation and classification problems can be increased. Binning and cardinality reduction are very similar procedures. They differ only in the fact that in the case of binning the order of the values is taken into account. Methods of binning and cardinality reduction range from simple ad hoc methods based on the understanding of the data and the experience of the analyst to more sophisticated methods relying on mathematical criteria. This chapter presents several of these methods—cardinality reduction, equal width binning, equal height binning, and optimal binning.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have