THIS PACKAGE HAS BEEN SUPERSEDED BY DISCRETIZERS.jl
This package supports discretization methods and mapping functions.
-
LabelMaps
map between discrete values -
BinMaps
map from continuous values to corresponding bins -
DataLabelMaps
andDataBinMaps
support NA values
Construct a binmap type mapping floats to Uint8 bin indeces:
bin_edges = [0.0, 0.5, 1.0, 2.0]
bmap = binmap(bin_edges, Uint8)
nlabels(bmap) -> 3
The primary interface is encoding and decoding:
encode(bmap, 0.25) -> 0x01
encode(bmap, [0.2, 1.5, 0.2]) -> [0x01, 0x02, 0x01]
decode(bmap, uint8(1)) -> rand(0.0:0.5)
BinMaps supports several binning algorithms, including:
- uniform width discretization
- uniform sample count discretization
- MODL bayes-optimal binning for continuous features over a discrete target
bin_edges = binedges(DISCRETIZE_UNIFORMWIDTH, 2, [0.0, 0.2, 0.1, 1.0, 0.6])
-> [0.0, 0.5, 1.0]
For questions please contact the package creator or create an issue
Feel free to create pull requests to improve the package