Difference between revisions of "Binned Variable"
From Open Risk Manual
Wiki admin (talk | contribs) |
Wiki admin (talk | contribs) |
||
Line 17: | Line 17: | ||
* [[Coarse Classification]] | * [[Coarse Classification]] | ||
− | [[Category: | + | [[Category:Probability]] |
Latest revision as of 13:49, 16 April 2021
Definition
A Binned Variable (also Grouped Variable) in the context of Quantitative Risk Management is any variable that is generated via the discretization of Numerical Variable into a defined set of bins (intervals).
Methods
- Uniform interval binning, which can be based on N subdivisions of the variable range, or some other approach
- Data adapted binning
- Unsupervised, with reference to the statistical distribution of the variable itself (for example Quantile Based Binning)
- Supervised, with reference to the statistical distribution of other variables
Examples
- Binned variables can be used in the representation of a Risk Model, encoding Explanatory Variables or a Risk Factor
- Binning of a continuous Credit Score into a small set of Credit Rating classes
Issues and Challenges
- Binning improves usability for some datasets, which needs to be offset against the potential loss of information through binning