Package 'FSelector'

Title: Selecting Attributes
Description: Functions for selecting attributes from a given dataset. Attribute subset selection is the process of identifying and removing as much of the irrelevant and redundant information as possible.
Authors: Piotr Romanski, Lars Kotthoff, Patrick Schratz
Maintainer: Lars Kotthoff <[email protected]>
License: GPL-2
Version: 0.34
Built: 2024-08-16 02:44:00 UTC
Source: https://github.com/larskotthoff/fselector

Help Index

Package for selecting attributes

Description

Package containing functions for selecting attributes from a given dataset and a destination attribute.

Details

This package contains:

  • -Algorithms for filtering attributes: cfs, chi.squared, information.gain, gain.ratio, symmetrical.uncertainty, linear.correlation, rank.correlation, oneR, relief, consistency, random.forest.importance

  • -Algorithms for wrapping classifiers and search attribute subset space: best.first.search, backward.search, forward.search, hill.climbing.search

  • -Algorithm for choosing a subset of attributes based on attributes' weights: cutoff.k, cutoff.k.percent, cutoff.biggest.diff

  • -Algorithm for creating formulas: as.simple.formula

Author(s)

Piotr Romanski
Maintainer: Lars Kotthoff <[email protected]>

Converting to formulas

Description

Converts character vector of atrributes' names and destination attribute's name to a simple formula.

Usage

as.simple.formula(attributes, class)

Arguments

attributes

character vector of attributes' names

class

name of destination attribute

Value

A simple formula like "class ~ attr1 + attr2"

Author(s)

Piotr Romanski

Examples

data(iris)
  result <- cfs(Species ~ ., iris)
  f <- as.simple.formula(result, "Species")

Best-first search

Description

The algorithm for searching atrribute subset space.

Usage

best.first.search(attributes, eval.fun, max.backtracks = 5)

Arguments

attributes

a character vector of all attributes to search in

eval.fun

a function taking as first parameter a character vector of all attributes and returning a numeric indicating how important a given subset is

max.backtracks

an integer indicating a maximum allowed number of backtracks, default is 5

Details

The algorithm is similar to forward.search besides the fact that is chooses the best node from all already evaluated ones and evaluates it. The selection of the best node is repeated approximately max.backtracks times in case no better node found.

Value

A character vector of selected attributes.

Author(s)

Piotr Romanski

See Also

forward.search, backward.search, hill.climbing.search, exhaustive.search

Examples

library(rpart)
  data(iris)
  
  evaluator <- function(subset) {
    #k-fold cross validation
    k <- 5
    splits <- runif(nrow(iris))
    results = sapply(1:k, function(i) {
      test.idx <- (splits >= (i - 1) / k) & (splits < i / k)
      train.idx <- !test.idx
      test <- iris[test.idx, , drop=FALSE]
      train <- iris[train.idx, , drop=FALSE]
      tree <- rpart(as.simple.formula(subset, "Species"), train)
      error.rate = sum(test$Species != predict(tree, test, type="c")) / nrow(test)
      return(1 - error.rate)
    })
    print(subset)
    print(mean(results))
    return(mean(results))
  }
  
  subset <- best.first.search(names(iris)[-5], evaluator)
  f <- as.simple.formula(subset, "Species")
  print(f)

CFS filter

Description

The algorithm finds attribute subset using correlation and entropy measures for continous and discrete data.

Usage

cfs(formula, data)

Arguments

formula

a symbolic description of a model

data

data to process

Details

The alorithm makes use of best.first.search for searching the attribute subset space.

Value

a character vector containing chosen attributes

Author(s)

Piotr Romanski

See Also

best.first.search

Examples

data(iris)
  
  subset <- cfs(Species~., iris)
  f <- as.simple.formula(subset, "Species")
  print(f)

Chi-squared filter

Description

The algorithm finds weights of discrete attributes basing on a chi-squared test.

Usage

chi.squared(formula, data)

Arguments

formula

a symbolic description of a model

data

a symbolic description of a model

Details

The result is equal to Cramer's V coefficient between source attributes and destination attribute.

Value

a data.frame containing the worth of attributes in the first column and their names as row names

Author(s)

Piotr Romanski

Examples

library(mlbench)
  data(HouseVotes84)

  weights <- chi.squared(Class~., HouseVotes84)
  print(weights)
  subset <- cutoff.k(weights, 5)
  f <- as.simple.formula(subset, "Class")
  print(f)

Consistency-based filter

Description

The algorithm finds attribute subset using consistency measure for continous and discrete data.

Usage

consistency(formula, data)

Arguments

formula

a symbolic description of a model

data

data to process

Details

The alorithm makes use of best.first.search for searching the attribute subset space.

Value

a character vector containing chosen attributes

Author(s)

Piotr Romanski

See Also

best.first.search

Examples

## Not run: 
  library(mlbench)
  data(HouseVotes84)
  
  subset <- consistency(Class~., HouseVotes84)
  f <- as.simple.formula(subset, "Class")
  print(f)

## End(Not run)

Correlation filter

Description

The algorithm finds weights of continous attributes basing on their correlation with continous class attribute.

Usage

linear.correlation(formula, data)
rank.correlation(formula, data)

Arguments

formula

a symbolic description of a model

data

data to process

Details

linear.correlation uses Pearson's correlation

rank.correlation uses Spearman's correlation

Rows with NA values are not taken into consideration.

Value

a data.frame containing the worth of attributes in the first column and their names as row names

Author(s)

Piotr Romanski

Examples

library(mlbench)
  data(BostonHousing)
  d=BostonHousing[-4] # only numeric variables
  
  weights <- linear.correlation(medv~., d)
  print(weights)
  subset <- cutoff.k(weights, 3)
  f <- as.simple.formula(subset, "medv")
  print(f)

  weights <- rank.correlation(medv~., d)
  print(weights)
  subset <- cutoff.k(weights, 3)
  f <- as.simple.formula(subset, "medv")
  print(f)

Cutoffs

Description

The algorithms select a subset from a ranked attributes.

Usage

cutoff.k(attrs, k)
cutoff.k.percent(attrs, k)
cutoff.biggest.diff(attrs)

Arguments

attrs

a data.frame containing ranks for attributes in the first column and their names as row names

k

a positive integer in case of cutoff.k and a numeric between 0 and 1 in case of cutoff.k.percent

Details

cutoff.k chooses k best attributes

cutoff.k.percent chooses best k * 100% of attributes

cutoff.biggest.diff chooses a subset of attributes which are significantly better than other.

Value

A character vector containing selected attributes.

Author(s)

Piotr Romanski

Examples

data(iris)

  weights <- information.gain(Species~., iris)
  print(weights)

  subset <- cutoff.k(weights, 1)
  f <- as.simple.formula(subset, "Species")
  print(f)

  subset <- cutoff.k.percent(weights, 0.75)
  f <- as.simple.formula(subset, "Species")
  print(f)

  subset <- cutoff.biggest.diff(weights)
  f <- as.simple.formula(subset, "Species")
  print(f)

Entropy-based filters

Description

The algorithms find weights of discrete attributes basing on their correlation with continous class attribute.

Usage

information.gain(formula, data, unit)
gain.ratio(formula, data, unit)
symmetrical.uncertainty(formula, data, unit)

Arguments

formula

A symbolic description of a model.

data

Data to process.

unit

Unit for computing entropy (passed to entropy. Default is "log".

Details

information.gain is

H(Class)+H(Attribute)−H(Class,Attribute)H(Class) + H(Attribute) - H(Class, Attribute)

.

gain.ratio is

H(Class)+H(Attribute)−H(Class,Attribute)H(Attribute)\frac{H(Class) + H(Attribute) - H(Class, Attribute)}{H(Attribute)}

symmetrical.uncertainty is

2H(Class)+H(Attribute)−H(Class,Attribute)H(Attribute)+H(Class)2\frac{H(Class) + H(Attribute) - H(Class, Attribute)}{H(Attribute) + H(Class)}

Value

a data.frame containing the worth of attributes in the first column and their names as row names

Author(s)

Piotr Romanski, Lars Kotthoff

Examples

data(iris)

  weights <- information.gain(Species~., iris)
  print(weights)
  subset <- cutoff.k(weights, 2)
  f <- as.simple.formula(subset, "Species")
  print(f)

  weights <- information.gain(Species~., iris, unit = "log2")
  print(weights)

  weights <- gain.ratio(Species~., iris)
  print(weights)
  subset <- cutoff.k(weights, 2)
  f <- as.simple.formula(subset, "Species")
  print(f)

  weights <- symmetrical.uncertainty(Species~., iris)
  print(weights)
  subset <- cutoff.biggest.diff(weights)
  f <- as.simple.formula(subset, "Species")
  print(f)

Exhaustive search

Description

The algorithm for searching atrribute subset space.

Usage

exhaustive.search(attributes, eval.fun)

Arguments

attributes

a character vector of all attributes to search in

eval.fun

a function taking as first parameter a character vector of all attributes and returning a numeric indicating how important a given subset is

Details

The algorithm searches the whole attribute subset space in breadth-first order.

Value

A character vector of selected attributes.

Author(s)

Piotr Romanski

See Also

forward.search, backward.search, best.first.search, hill.climbing.search

Examples

library(rpart)
  data(iris)
  
  evaluator <- function(subset) {
    #k-fold cross validation
    k <- 5
    splits <- runif(nrow(iris))
    results = sapply(1:k, function(i) {
      test.idx <- (splits >= (i - 1) / k) & (splits < i / k)
      train.idx <- !test.idx
      test <- iris[test.idx, , drop=FALSE]
      train <- iris[train.idx, , drop=FALSE]
      tree <- rpart(as.simple.formula(subset, "Species"), train)
      error.rate = sum(test$Species != predict(tree, test, type="c")) / nrow(test)
      return(1 - error.rate)
    })
    print(subset)
    print(mean(results))
    return(mean(results))
  }
  
  subset <- exhaustive.search(names(iris)[-5], evaluator)
  f <- as.simple.formula(subset, "Species")
  print(f)

Greedy search

Description

The algorithms for searching atrribute subset space.

Usage

backward.search(attributes, eval.fun)
forward.search(attributes, eval.fun)

Arguments

attributes

a character vector of all attributes to search in

eval.fun

a function taking as first parameter a character vector of all attributes and returning a numeric indicating how important a given subset is

Details

These algorithms implement greedy search. At first, the algorithms expand starting node, evaluate its children and choose the best one which becomes a new starting node. This process goes only in one direction. forward.search starts from an empty and backward.search from a full set of attributes.

Value

A character vector of selected attributes.

Author(s)

Piotr Romanski

See Also

best.first.search, hill.climbing.search, exhaustive.search

Examples

library(rpart)
  data(iris)
  
  evaluator <- function(subset) {
    #k-fold cross validation
    k <- 5
    splits <- runif(nrow(iris))
    results = sapply(1:k, function(i) {
      test.idx <- (splits >= (i - 1) / k) & (splits < i / k)
      train.idx <- !test.idx
      test <- iris[test.idx, , drop=FALSE]
      train <- iris[train.idx, , drop=FALSE]
      tree <- rpart(as.simple.formula(subset, "Species"), train)
      error.rate = sum(test$Species != predict(tree, test, type="c")) / nrow(test)
      return(1 - error.rate)
    })
    print(subset)
    print(mean(results))
    return(mean(results))
  }
  
  subset <- forward.search(names(iris)[-5], evaluator)
  f <- as.simple.formula(subset, "Species")
  print(f)

Hill climbing search

Description

The algorithm for searching atrribute subset space.

Usage

hill.climbing.search(attributes, eval.fun)

Arguments

attributes

a character vector of all attributes to search in

eval.fun

a function taking as first parameter a character vector of all attributes and returning a numeric indicating how important a given subset is

Details

The algorithm starts with a random attribute set. Then it evaluates all its neighbours and chooses the best one. It might be susceptible to local maximum.

Value

A character vector of selected attributes.

Author(s)

Piotr Romanski

See Also

forward.search, backward.search, best.first.search, exhaustive.search

Examples

library(rpart)
  data(iris)
  
  evaluator <- function(subset) {
    #k-fold cross validation
    k <- 5
    splits <- runif(nrow(iris))
    results = sapply(1:k, function(i) {
      test.idx <- (splits >= (i - 1) / k) & (splits < i / k)
      train.idx <- !test.idx
      test <- iris[test.idx, , drop=FALSE]
      train <- iris[train.idx, , drop=FALSE]
      tree <- rpart(as.simple.formula(subset, "Species"), train)
      error.rate = sum(test$Species != predict(tree, test, type="c")) / nrow(test)
      return(1 - error.rate)
    })
    print(subset)
    print(mean(results))
    return(mean(results))
  }
  
  subset <- hill.climbing.search(names(iris)[-5], evaluator)
  f <- as.simple.formula(subset, "Species")
  print(f)

OneR algorithm

Description

The algorithms find weights of discrete attributes basing on very simple association rules involving only one attribute in condition part.

Usage

oneR(formula, data)

Arguments

formula

a symbolic description of a model

data

data to process

Details

The algorithm uses OneR classifier to find out the attributes' weights. For each attribute it creates a simple rule based only on that attribute and then calculates its error rate.

Value

a data.frame containing the worth of attributes in the first column and their names as row names

Author(s)

Piotr Romanski

Examples

library(mlbench)
  data(HouseVotes84)
  
  weights <- oneR(Class~., HouseVotes84)
  print(weights)
  subset <- cutoff.k(weights, 5)
  f <- as.simple.formula(subset, "Class")
  print(f)

RandomForest filter

Description

The algorithm finds weights of attributes using RandomForest algorithm.

Usage

random.forest.importance(formula, data, importance.type = 1)

Arguments

formula

a symbolic description of a model

data

data to process

importance.type

either 1 or 2, specifying the type of importance measure (1=mean decrease in accuracy, 2=mean decrease in node impurity)

Details

This is a wrapper for importance.

Value

a data.frame containing the worth of attributes in the first column and their names as row names

Author(s)

Piotr Romanski

Examples

library(mlbench)
  data(HouseVotes84)
  
  weights <- random.forest.importance(Class~., HouseVotes84, importance.type = 1)
  print(weights)
  subset <- cutoff.k(weights, 5)
  f <- as.simple.formula(subset, "Class")
  print(f)

RReliefF filter

Description

The algorithm finds weights of continous and discrete attributes basing on a distance between instances.

Usage

relief(formula, data, neighbours.count = 5, sample.size = 10)

Arguments

formula

a symbolic description of a model

data

data to process

neighbours.count

number of neighbours to find for every sampled instance

sample.size

number of instances to sample

Details

The algorithm samples instances and finds their nearest hits and misses. Considering that result, it evaluates weights of attributes.

Value

a data.frame containing the worth of attributes in the first column and their names as row names

Author(s)

Piotr Romanski

References

  • -Igor Kononenko: Estimating Attributes: Analysis and Extensions of RELIEF. In: European Conference on Machine Learning, 171-182, 1994.

  • -Marko Robnik-Sikonja, Igor Kononenko: An adaptation of Relief for attribute estimation in regression. In: Fourteenth International Conference on Machine Learning, 296-304, 1997.

Examples

data(iris)
  
  weights <- relief(Species~., iris, neighbours.count = 5, sample.size = 20)
  print(weights)
  subset <- cutoff.k(weights, 2)
  f <- as.simple.formula(subset, "Species")
  print(f)