Skip to content
a Ruby gem for feature selection and ranking
Ruby JavaScript
Find file
Latest commit d4e96ee Jan 4, 2013 @need47 update gemspec file
Failed to load latest commit information.
doc update to version 1.4.0 Nov 5, 2012
lib update to version 1.4.0 Nov 5, 2012
ChangeLog update to version 1.4.0 Nov 5, 2012
HowToContribute correct spell error May 31, 2012
LICENSE LICENSE Apr 2, 2012 update to version 1.4.0 Nov 5, 2012
Rakefile initial package import Feb 27, 2012

FSelector: a Ruby gem for feature selection

Source Code:
Publication: Bioinformatics, 2012, 28, 2851-2852
Author: Tiejun Cheng
Copyright: 2012
License: MIT License
Latest Version: 1.4.0
Release Date: 2012-11-05


FSelector is a Ruby gem that aims to integrate various feature selection algorithms and related functions into one single package. Welcome to contact me ( if you'd like to contribute your own algorithms or report a bug. FSelector allows user to perform feature selection by using either a single algorithm or an ensemble of multiple algorithms, and other common tasks including normalization and discretization on continuous data, as well as replace missing feature values with certain criterion. FSelector acts on a full-feature data set in either CSV, LibSVM or WEKA file format and outputs a reduced data set with only selected subset of features, which can later be used as the input for various machine learning softwares such as LibSVM and WEKA. FSelector, as a collection of filter methods, does not implement any classifier like support vector machines or random forest. Check below for a list of FSelector's features, {file:ChangeLog} for updates, and {file:HowToContribute} if you want to contribute.

Feature List

1. supported input/output file types

  • csv
  • libsvm
  • weka ARFF
  • on-line dataset in one of the above three formats (read only)
  • random data (read only, for test purpose)

2. available feature selection/ranking algorithms

algorithm                        shortcut    algo_type  applicability  feature_type
Accuracy                         Acc         weighting  multi-class    discrete
AccuracyBalanced                 Acc2        weighting  multi-class    discrete
BiNormalSeparation               BNS         weighting  multi-class    discrete
CFS_d                            CFS_d       searching  multi-class    discrete
ChiSquaredTest                   CHI         weighting  multi-class    discrete
CorrelationCoefficient           CC          weighting  multi-class    discrete
DocumentFrequency                DF          weighting  multi-class    discrete
F1Measure                        F1          weighting  multi-class    discrete
FishersExactTest                 FET         weighting  multi-class    discrete
FastCorrelationBasedFilter       FCBF        searching  multi-class    discrete
GiniIndex                        GI          weighting  multi-class    discrete
GMean                            GM          weighting  multi-class    discrete
GSSCoefficient                   GSS         weighting  multi-class    discrete
InformationGain                  IG          weighting  multi-class    discrete
INTERACT                         INTERACT    searching  multi-class    discrete
JMeasure                         JM          weighting  multi-class    discrete
KLDivergence                     KLD         weighting  multi-class    discrete
MatthewsCorrelationCoefficient   MCC, PHI    weighting  multi-class    discrete
McNemarsTest                     MNT         weighting  multi-class    discrete
OddsRatio                        OR          weighting  multi-class    discrete
OddsRatioNumerator               ORN         weighting  multi-class    discrete
PhiCoefficient                   PHI         weighting  multi-class    discrete
Power                            Power       weighting  multi-class    discrete
Precision                        Precision   weighting  multi-class    discrete
ProbabilityRatio                 PR          weighting  multi-class    discrete
Recall                           Recall      weighting  multi-class    discrete
Relief_d                         Relief_d    weighting  two-class      discrete
ReliefF_d                        ReliefF_d   weighting  multi-class    discrete
Sensitivity                      SN, Recall  weighting  multi-class    discrete
Specificity                      SP          weighting  multi-class    discrete
SymmetricalUncertainty           SU          weighting  multi-class    discrete
BetweenWithinClassesSumOfSquare  BSS_WSS     weighting  multi-class    continuous
CFS_c                            CFS_c       searching  multi-class    continuous
FTest                            FT          weighting  multi-class    continuous
KS_CCBF                          KS_CCBF     searching  multi-class    continuous
KSTest                           KST         weighting  two-class      continuous
PMetric                          PM          weighting  two-class      continuous
Relief_c                         Relief_c    weighting  two-class      continuous
ReliefF_c                        ReliefF_c   weighting  multi-class    continuous
TScore                           TS          weighting  two-class      continuous
WilcoxonRankSum                  WRS         weighting  two-class      continuous
LasVegasFilter                   LVF         searching  multi-class    discrete, continuous, mixed
LasVegasIncremental              LVI         searching  multi-class    discrete, continuous, mixed
Random                           Rand        weighting  multi-class    discrete, continuous, mixed
RandomSubset                     RandS       searching  multi-class    discrete, continuous, mixed

note for feature selection interface:
there are two types of filter algorithms: filter_by_feature_weighting and filter_by_feature_searching

  • for former: use either select_feature_by_score! or select_feature_by_rank!
  • for latter: use select_feature!

3. feature selection approaches

  • by a single algorithm
  • by multiple algorithms in a tandem manner
  • by multiple algorithms in an ensemble manner (share the same feature selection interface as single algorithm)

4. availabe normalization and discretization algorithms for continuous feature

algorithm                         note
normalize_by_log!                 normalize by logarithmic transformation
normalize_by_min_max!             normalize by scaling into [min, max]
normalize_by_zscore!              normalize by converting into zscore
discretize_by_equal_width!        discretize by equal width among intervals
discretize_by_equal_frequency!    discretize by equal frequency among intervals
discretize_by_ChiMerge!           discretize by ChiMerge algorithm
discretize_by_Chi2!               discretize by Chi2 algorithm
discretize_by_MID!                discretize by Multi-Interval Discretization algorithm
discretize_by_TID!                discretize by Three-Interval Discretization algorithm

5. availabe algorithms for replacing missing feature values

algorithm                         note                                   feature_type                     
replace_by_fixed_value!           replace by a fixed value               discrete, continuous
replace_by_mean_value!            replace by mean feature value          continuous
replace_by_median_value!          replace by median feature value        continuous
replace_by_knn_value!             replace by weighted knn feature value  continuous
replace_by_most_seen_value!       replace by most seen feature value     discrete


To install FSelector, use the following command:

$ gem install fselector

note: From version 0.5.0, FSelector uses the RinRuby gem ( as a seemless bridge to access the statistical routines in the R package (, which will greatly expand the inclusion of algorithms to FSelector, especially for those relying on statistical test. To this end, please pre-install the R package. RinRuby should have been auto-installed with FSelector via the above command.


1. feature selection by a single algorithm

require 'fselector'

# use InformationGain (IG) as a feature selection algorithm
r1 =

# read from random data (or csv, libsvm, weka ARFF file)
# no. of samples: 100
# no. of classes: 2
# no. of features: 15
# no. of possible values for each feature: 3
# allow missing values: true
r1.data_from_random(100, 2, 15, 3, true)

# number of features before feature selection
puts "  # features (before): "+ r1.get_features.size.to_s

# select the top-ranked features with scores >0.01

# number of features after feature selection
puts "  # features (after): "+ r1.get_features.size.to_s

# you can also use a second alogirithm for further feature selection
# e.g. use the ChiSquaredTest (CHI) with Yates' continuity correction
# initialize from r1's data
r2 =, r1.get_data)

# number of features before feature selection
puts "  # features (before): "+ r2.get_features.size.to_s

# select the top-ranked 3 features

# number of features after feature selection
puts "  # features (after): "+ r2.get_features.size.to_s

# save data to standard ouput as a weka ARFF file (sparse format)
# with selected features only
r2.data_to_weka(:stdout, :sparse)

2. feature selection by an ensemble of multiple feature selectors

require 'fselector'

# example 1

# creating an ensemble of feature selectors by using 
# a single feature selection algorithm (INTERACT) 
# by instance perturbation (e.g. random sampling)

# test for the type of feature subset selection algorithms
r =

# an ensemble of 40 feature selectors with 90% data by random sampling
re =, 40, 0.90, :random_sampling)

# read SPECT data set (under the test/ directory)

# number of features before feature selection
puts '  # features (before): ' + re.get_features.size.to_s

# only features with above average count among ensemble are selected

# number of features after feature selection
puts '  # features before (after): ' + re.get_features.size.to_s

# example 2

# creating an ensemble of feature selectors by using 
# two feature selection algorithms: InformationGain (IG) and Relief_d. 
# note: can be 2+ algorithms, as long as they are of the same type, 
# either filter_by_feature_weighting or filter_by_feature_searching

# test for the type of feature weighting algorithms 
r1 =
r2 =

# an ensemble of two feature selectors
re =, r2)

# read random discrete data (containing missing value)
re.data_from_random(100, 2, 15, 3, true)

# replace missing value because Relief_d 
# does not allow missing value

# number of features before feature selection
puts '  # features (before): ' + re.get_features.size.to_s

# based on the max feature score (z-score standardized) among
# an ensemble of feature selectors
re.ensemble_by_score(:by_max, :by_zscore)

# select the top-ranked 3 features

# number of features after feature selection
puts '  # features (after): ' + re.get_features.size.to_s

3. feature selection after discretization

require 'fselector'

# the Information Gain (IG) algorithm requires data with discrete feature
r =

# but the Iris data set contains continuous features
r.data_from_url('', :weka)

# let's first discretize it by ChiMerge algorithm at alpha=0.10
# then perform feature selection as usual

# number of features before feature selection
puts '  # features (before): ' + r.get_features.size.to_s

# select the top-ranked feature

# number of features after feature selection
puts '  # features (after): ' + r.get_features.size.to_s

4. see more examples test_*.rb under the test/ directory

How to contribute

check {file:HowToContribute} to see how to write your own feature selection algorithms and/or make contribution to FSelector.

Change Log

A {file:ChangeLog} is available from version 0.5.0 and upward to refelect what's new and what's changed.


FSelector © 2012 by Tiejun Cheng. FSelector is licensed under the MIT license. Please see the {file:LICENSE} for more information.

Something went wrong with that request. Please try again.