Description Usage Arguments Details Value Author(s) References See Also Examples
The method discretize
returns discretization bounds for numeric attributes and two auxiliary functions.
Discretization can be obtained with one of the three discretization methods:
greedy search using given feature evaluation heuristics, equal width of intervals, or equal number of instances in each interval.
The attributes and target variable are specified using formula interface, target variable name or index.
Feature evaluation algorithms available for classification problems
are various variants of Relief and ReliefF algorithms, gain ratio, giniindex, MDL, DKM, information gain, etc.
For regression problems there are RREliefF, MSEofMean, MSEofModel, MAEofMode, etc.
1 2 3 4 5 6 7 8 9  discretize(formula, data, method=c("greedy", "equalFrequency", "equalWidth"),
estimator, discretizationLookahead=3, discretizationSample=0,
maxBins=0, equalDiscBins=4, ...)
applyDiscretization(data, boundsList, noDecimalsInValueName=2)
intervalMidPoint(data, boundsList,
midPointMethod=c("equalFrequency", "equalWidth"))

formula 
Either a formula specifying the attributes to be evaluated and the target variable, or a name of target variable, or an index of target variable. 
data 
Data frame with data. 
method 
Three discretization methods are available. With 
estimator 
The name of the evaluation method. 
discretizationLookahead 
Discretization is performed with a greedy algorithm which adds a new boundary, until there is no
improvement in evaluation function for 
discretizationSample 
Maximal number of points to try discretization (0=all sensible). Binarization of multivalued discrete features with
k values is performed exhaustively, if 2^k  1 is at most 
maxBins 
The maximal number of discrete bins for numeric attributes used for greedy discretization (0=don't care).
This shall be an integer vector of length
equal to the number of numeric attributes or an integer which applies to all numeric attributes. The default value of
0 means that the number of bins will be determined greedily taking into account 
.
equalDiscBins 
The number of bins used in equal frequency and equal width discretization. This shall be an integer vector of length equal to the number of numeric attributes or an integer which applies to all numeric attributes. The default value is 4. 
... 
Additional options used by specific evaluation methods as described in 
boundsList 
A list of numeric bounds which is applied to numeric attributes in 
.
noDecimalsInValueName 
With how many decimal places will the numeric feature values be presented in description (i.e., levels) of feature values. The default value is 2, but will be increased if this is necessary to avoid the same description of feature values. 
midPointMethod 
Two methods to determine the middle points of discretization intervals are available.
The 
In method discretize
the parameter formula
can be interpreted in three ways, where the formula interface is the most elegant one,
but inefficient and inappropriate for large data sets. See CoreModel
for details.
The estimator parameter selects the evaluation heuristics. For classification problem it
must be one of the names returned by infoCore(what="attrEval")
and for
regression problem it must be one of the names returned by infoCore(what="attrEvalReg")
.
For details see their description in attrEval
.
If the number of supplied vector in maxBins
and equalDiscBins
is shorter than the number of numeric attributes, the
vector is coerced to the required length.
There are some additional parameters ... available which are used by specific evaluation heuristics.
Their list and short description is available by calling helpCore
. See Section on attribute evaluation.
The function applyDiscretization
takes the discretization bounds obtain with function discretize
and transforms
numeric features in a data set into discrete features.
The function intervalMidPoint
takes discretization bounds provided by function discretize
and returns
middle points of discretization intervals for numeric attributes. The middle points are computed from the data;
for lowest/highest interval the minimum/maximum of the values in the data
for particular attribute
are implicitly taken as an additional left/right boundary point.
The method discretize
returns a list of discretization bounds for numeric attributes. One component of a list contains bounds for one attribute.
If an attribute has all values equal, value NA is returned. If an attribute has all values equal to NA, it is skipped in the returned list.
The function applyDiscretization
returns a data set where all numeric attributes are replaced with their discrete versions.
The function intervalMidPoint
returns a list of vectors where each vector contains middle point of discretized intevals.
Marko RobnikSikonja
Marko RobnikSikonja, Igor Kononenko: Theoretical and Empirical Analysis of ReliefF and RReliefF. Machine Learning Journal, 53:2369, 2003
Marko RobnikSikonja, Igor Kononenko: Discretization of continuous attributes using ReliefF. Proceedings of ERK'95 , Portoroz, Slovenia, 1995.
Some of these references are available also from http://lkm.fri.unilj.si/rmarko/papers/
CORElearn
,
CoreModel
,
attrEval
,
helpCore
,
infoCore
.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16  # use iris data
# run method using estimator ReliefF with exponential rank distance
discBounds < discretize(Species ~ ., iris, method="greedy",
estimator="ReliefFexpRank")
print(discBounds)
discreteIris < applyDiscretization(iris, discBounds)
prototypePoints < intervalMidPoint(iris, discBounds,
midPointMethod="equalFrequency")
regData < regDataGen(200)
discretize(response ~ ., regData, method="greedy", estimator="RReliefFequalK",
maxBins=2)
# print all available estimators
#infoCore(what="attrEval")
#infoCore(what="attrEvalReg")

Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.