Title: The MCFS-ID Algorithm for Feature Selection and Interdependency Discovery
Version: 1.3.6
Date: 2024-08-18
Depends: rJava (≥ 0.5-0), R (≥ 2.70)
Suggests: testthat, R.rsp
VignetteBuilder: R.rsp
Imports: yaml, ggplot2, gridExtra, reshape2, dplyr, stringi, igraph (≥ 2.0.0), data.table (≥ 1.0.1)
SystemRequirements: Java (>= 7)
Description: MCFS-ID (Monte Carlo Feature Selection and Interdependency Discovery) is a Monte Carlo method-based tool for feature selection. It also allows for the discovery of interdependencies between the relevant features. MCFS-ID is particularly suitable for the analysis of high-dimensional, 'small n large p' transactional and biological data. M. Draminski, J. Koronacki (2018) <doi:10.18637/jss.v085.i12>.
License: GPL-3
URL: https://home.ipipan.waw.pl/m.draminski/mcfs.html
NeedsCompilation: no
Packaged: 2024-08-19 11:44:48 UTC; mdraminski
Author: Michal Draminski [aut, cre], Jacek Koronacki [aut], Julian Zubek [ctb]
Maintainer: Michal Draminski <michal.draminski@ipipan.waw.pl>
Repository: CRAN
Date/Publication: 2024-08-19 14:40:02 UTC

Creates artificial dataset

Description

Creates data.frame with artificial data. The last six columns are nominal and highly correlated to feature 'class'. This data set consists of objects from 3 classes, A, B and C, that contain 40, 20, 10 objects, respectively (70 objects altogether). For each object, 6 binary features (A1, A2, B1, B2, C1 and C2) are created and they are 'ideally' or 'almost ideally' correlated with class feature. If an object's 'class' equals 'A', then its features A1 and A2 are set to class value 'A'; otherwise A1 = A2 = 0. If an object's 'class' is 'B' or 'C', the processing is analogous, but some random corruption is introduced. For 2 observations from class 'B' and both attributes B1/B2, their values 'B' are replaced by '0'. For 4 observations from class 'C' and both attributes C1/C2, their values 'C' are replaced by '0'. The number of corrupted values for each class is defined by corruption parameter. The data also contains additional rnd_features = 500 random numerical features with uniformly [0,1] distributed values.

Usage

artificial.data(rnd_features = 500, size = c(40, 20, 10), 
                        corruption = c(0, 2, 4), seed = NA)

Arguments

rnd_features

number of numerical random features.

size

size of classes A, B, and C.

corruption

defines the number of corrupted values for a pairs of columns A1/A2, B1/B2, C1/C2,

seed

seed for random number generator.

Value

data.frame with six important features.

Examples

  d <- artificial.data(rnd_features = 500)
  showme(d)

Constructs interdependencies graph

Description

Constructs the ID-Graph (igraph/idgraph object) from mcfs_result object returned by mcfs function. The number of top features included and the number of ID-Graph edges can be customized.

Usage

build.idgraph(mcfs_result, 
                      size = NA, 
                      size_ID = NA, 
                      self_ID = FALSE,
                      outer_ID = FALSE,
                      orphan_nodes = FALSE, 
                      size_ID_mult = 3, 
                      size_ID_max = 100)

Arguments

mcfs_result

results returned by mcfs function.

size

number of top features to select. If size = NA, then size is defined by mcfs_result$cutoff_value parameter.

size_ID

number of interdependencies (edges in ID-Graph) to be included. If size_ID = NA, then parameter size_ID is defined by multiplication size_ID_mult*size.

self_ID

if self_ID = TRUE, then include self-loops from ID-Graph.

outer_ID

if outer_ID = TRUE, then include include all interactions between a feature from the top set features (defined by size parameter) with any other feature.

orphan_nodes

if plot_all_nodes = TRUE, then include all nodes, even if they are not connected to any other node (isolated nodes).

size_ID_mult

If size_ID_mult = 3 there will be 3 times more edges than features (nodes) presented on the ID-Graph. It works only if size = NA and size_ID = NA

size_ID_max

maximum number of interactions to be included from ID-Graph (the upper limit).

Value

igraph/idgraph S3 object that can be: plotted in R, exported to graphML (XML format) or saved as csv or rds files.

Examples

  ## Not run: ###dontrunbegin

  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 50,
                  buildID = TRUE, finalCV = FALSE, finalRuleset = FALSE,
                  threadsNumber = 2)

  # build interdependencies graph for top 6 features 
  # and top 12 interdependencies and plot all nodes
  gid <- build.idgraph(result, size = 6, size_ID = 12, orphan_nodes = TRUE)
  plot(gid, label_dist = 1)

  # Export graph to graphML (XML structure)
  path <- tempdir()
  igraph::write_graph(gid, file = file.path(path, "artificial.graphml"), 
            format = "graphml", prefixAttr = FALSE)

  
## End(Not run)###dontrunend

Exports MCFS-ID result plots

Description

Saves all MCFS-ID result plots in the specified directory.

Usage

export.plots(mcfs_result, data = NULL, idgraph = NULL, 
                    path, label = "mcfs", color = "darkred",
                    size = NA, image_width = 8, image_height = 6, 
                    plot_format = c("pdf","svg","png"), cex = 1)

Arguments

mcfs_result

result from mcfs function.

data

input data frame used to produce mcfs_result.

idgraph

idgraph/igraph S3 object representing feature interdependencies. This object is produced by build.idgraph function.

path

path to the where plot files should be saved.

label

a common prefix label of all plot files.

color

it defines main color of all plots.

size

number of features to plot.

image_width

width of plots (in inches).

image_height

height of plots (in inches).

plot_format

image format of plot files - one of the following: "pdf","svg","png".

cex

size of fonts.

Examples

  ## Not run: ###dontrunbegin
  
  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 50,
                  finalCV = FALSE, finalRuleset = FALSE, threadsNumber = 2)
  
  # build interdependencies graph for top 6 features 
  # and top 12 interdependencies and plot all nodes
  gid <- build.idgraph(result, size = 6, size_ID = 12, orphan_nodes = TRUE)

  #export plot files
  export.plots(result, adata, idgraph = gid, path = tempdir(), label = "mcfs", color = "darkgreen")

  
## End(Not run)###dontrunend

Saves MCFS-ID result into set csv files

Description

Saves csv files with result obtained by the MCFS-ID.

Usage

export.result(mcfs_result, path = "./", label = "rmcfs", zip = TRUE)

Arguments

mcfs_result

result of the MCFS-ID experiment returned by mcfs function.

path

path to the MCFS-ID results files. This parameter can also point to the zip result file.

label

label of the experiment and common name for output files.

zip

if = TRUE, saves all results data as one zip file.

Examples

  ## Not run: ###dontrunbegin
  
  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 10,
                  finalCV = FALSE, finalRuleset = FALSE, threadsNumber = 2)

  # Export and import R result to/from files
  path <- file.path(tempdir(), "artificial.zip")
  export.result(result, path = path)
  result <- import.result(path = path)
  
  
## End(Not run)###dontrunend

Fixes input data values, column names and attributes types

Description

Fixes any input data to prepare them to export to ARFF/ADX formats. If after exporting data to ARFF/ADX formats there are some problems in running Java MCFS or WEKA, try to use this function before. This function fixes data values (e.g. space " " is replaced by "_") and data types (e.g. all Date columns converted to character in R).

Usage

fix.data(x, 
          type = c("all", "names", "values", "types"), 
          source_chars = c(" ", "'", ",", "/", "|", "#", 
                           "-", "(", ")", "[", "]", "{", "}"),
          destination_char = "_", 
          numeric_class = c("difftime"), 
          nominal_class = c("factor", "logical", "Date", "POSIXct", "POSIXt"))

Arguments

x

input data frame to be fixed.

type

  • all - fixes: column names, data values, data types.

  • names - fixes only column names. All characters determined by source_chars parameter are replaced by destination_char (e.g. space " " is replaced by "_").

  • values - fixes only data values. All characters determined by source_chars parameter are replaced by destination_char (e.g. space " " is replaced by "_").

  • types - fixes only data types (e.g. all possible nominal columns as (Date or logical) converted to character).

source_chars

characters that will be replaced in column names and data values.

destination_char

character that will be inserted in column names and data values.

numeric_class

vector of class labels to be casted as.numeric.

nominal_class

vector of class labels to be casted as.character.

Value

data.frame with fixed values and types (depends on type parameter).

Examples

  ## Not run: ###dontrunbegin

  # create artificial data
  adata <- artificial.data(rnd_features = 1000)
  
  # Fix data types and data values - remove "," " " "/" from values and fix data types
  # This function may help if mcfs has any problems with input data
  adata.fixed <- fix.data(adata)
  
  
## End(Not run)###dontrunend

Reads csv result files produced by the MCFS-ID Java module

Description

Reads csv result files produced by the MCFS-ID Java module.

Usage

import.result(path = "./", label = NA)

Arguments

path

path to the MCFS-ID results files. This parameter can also point to the zip result file.

label

experiment label for results files (name of the data).

Value

the result of the MCFS-ID experiment returned by mcfs function.

Examples

  ## Not run: ###dontrunbegin

  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 10,
                  finalCV = FALSE, finalRuleset = FALSE, threadsNumber = 2)

  # Export and import R result to/from files
  path <- file.path(tempdir(), "artificial.zip")
  export.result(result, path = path)
  result <- import.result(path = path)

  
  
## End(Not run)###dontrunend

MCFS-ID (Monte Carlo Feature Selection and Interdependency Discovery)

Description

Performs Monte Carlo Feature Selection (MCFS-ID) on a given data set. The data set should define a classification problem with discrete/nominal class labels. This function returns features sorted by RI as well as cutoff value, ID-Graph edges that denote interdependencies (ID), evaluation of top features and other statistics.

Usage

mcfs(formula, data,
    attrWeights = NULL,
    projections = 'auto',
    projectionSize = 'auto',
    featureFreq = 100,
    splits = 5,
    splitSetSize = 500,
    balance = 'auto',
    cutoffMethod = c("permutations", "criticalAngle", "kmeans", "mean", "contrast"),
    cutoffPermutations = 20,
    mode = 1,
    buildID = TRUE,
    finalRuleset = TRUE,
    finalCV = TRUE,
    finalCVSetSize = 1000,
    seed = NA,
    threadsNumber = 4)

Arguments

formula

specifies decision attribute and relation between class and other attributes (e.g. class~.). The target attribute can be nominal (then MCFS-ID uses decision tree) or numerical (then MCFS-ID uses regression tree).

data

defines input data.frame containing all features with decision attribute included. This data.frame must contain proper types of columns. Columns character, factor, Date, POSIXct, POSIXt are treated as nominal/categorical and remaining columns as numerical/continuous. Decision attribute defined by formula can be nominal or numerical.

attrWeights

defines vector of length = ncol(data) of attributes weights - weight 10 denotes 10 times larger chance for the attribute to be selected to the random subset than if weight equals to 1.

projections

defines the number of subsets (projections) with randomly selected features. This parameter is usually set to a few thousands and is denoted in the paper as s. By default it is set to 'auto' and this value is based on size of input data set and featureFreq parameter.

projectionSize

defines the number of features in one subset. It can be defined by an absolute value (e.g. 100 denotes 100 randomly selected features) or by a fraction of input attributes (e.g. 0.05 denotes 5% of input features). This parameter is denoted in the paper as m. If is set to 'auto' then projectionSize equals to \sqrt{d}, where d is the number of input features. Minimum number of features in one subset is 1.

featureFreq

determines how many times each input feature should be randomly selected when projections = 'auto'.

splits

defines the number of splits of each subset. This parameter is denoted in the paper as t. The size of the training set in the input subset is always set on 66%.

splitSetSize

determines whether to limit input dataset size. It helps to speedup computation for data sets with a large number of objects. If the parameter is larger than 1, it determines the number of objects that are drawn at random for each of the s \cdot t decision trees. If splitSetSize = 0 then the MCFS uses all objects in each iteration.

balance

determines the way to balance classes. It should be set to 2 or higher if input dataset contains heavily unbalanced classes. Each subset s will contain all the objects from the least frequent class and randomly selected set of objects from each of the remaining classes. This option helps to select features that are important for discovering a relatively rare class. The parameter defines the maximal imbalance ratio. If the ratio is set to 2, then subset s will contain the number of objects from each class (but the least frequent one) proportional to the square root of the class size size(c)^{1/2}. If balance = 0 then balancing is turned off. If balance = 1 it is on but does not change the size of classes. Default value is 'auto'.

cutoffMethod

determines the final cutoff method. Default value is 'permutations'. The methods of finding cutoff value between important and unimportant attributes are the following:

  • permutations - the method consists in permuting the decision attribute at least 20 times and running the MCFS-ID algorithm for each permutation. The set of the maximal RIs from all these experiments is assumed approximately normally distributed and a critical value based on the the one-sided (upper-tailed) Student's t-test (at 95% significance level) is provided. A feature is declared informative if its RI in the original ranking (without any permutation) exceeds the obtained critical value. A more detailed description of this method is included in the paper.

  • criticalAngle - critical angle method is based on the plot of the features' RIs in decreasing order of size, with the corresponding features equally spaced along the abscissa. The plot can be seen as piecewise linear function, where each linear segment joins two neighboring RIs. Roughly speaking, the cutoff (placed on the abscissa) corresponds to this point on the plot where the slope of consecutive segments changes significantly.

  • kmeans - the method is based on clustering the RI values into two clusters by the k-means algorithm. It sets the cutoff where the two clusters are separated. This method is quite valuable when data contains a subset of very informative features.

  • mean - cutoff value is set on mean values obtained from all the implemented methods.

  • contrast - This method adds 10% contrast (pure numerical random) atributes to the data then MCFS-ID is executed. Position of top 5% of them determines cutoff value. Usually it gives the largest cutoff beacause it select all attributes that are more informative than pure noise.

cutoffPermutations

determines the number of permutation runs. It needs at least 20 permutations (cutoffPermutations = 20) for a statistically significant result. Minimum value of this parameter is 3, however if it is 0 then permutations method is turned off.

mode

determines number of stages in MCFS filtering. If mode = 2 then MCFS is running new method that is based on two stage filtering. This method is much faster for BIG DATA - 1st stage filtering is performed based on contrast attributes (same as cutoffMethod = 'contrast') and 2nd stage is performed based on permutations experiments. If mode = 1 then it always runs one stage filtering the same as in rmcfs 1.2.x.

buildID

if = TRUE, Interdependencies Discovery is on and all ID-Graph edges are collected.

finalRuleset

if = TRUE, classification rules (by ripper algorithm) are created on the basis of the final set of features.

finalCV

if = TRUE, it runs 10 folds cross validation (cv) experiments on the final set of features. The following set of classifiers is used: C4.5, NB, SVM, kNN, logistic regression and Ripper.

finalCVSetSize

limits the number of objects used in the final cv experiment. For each out of 3 cv repetitions, the objects are selected randomly from the uniform distribution.

seed

seed for random number generator in Java. By default the seed is random. Replication of the result is possible only if threadsNumber = 1.

threadsNumber

number of threads to use in computation. More threads needs more CPU cores as well as memory usage is a bit higher. It is recommended to set this value equal to or less than CPU available cores.

Value

data

input data.frame limited to the top important features set.

target

decision attribute name.

RI

data.frame that contains all features with relevance scores sorted from the most relevant to the least relevant. This is the ranking of features.

ID

data.frame that contains features interdependencies as graph edges. It can be converted into a graph object by build.idgraph function.

distances

data.frame that contains convergence statistics of subsequent projections.

cmatrix

confusion matrix obtained from all s \cdot t decision trees.

cutoff

data.frame that contains cutoff values obtained by the following methods: mean, kmeans, criticalAngle, permutations (max RI).

cutoff_value

the number of features chosen as informative by the method defined by parameter cutoffMethod.

cv_accuracy

data.frame that contains classification results obtained by cross validation performed on cutoff_value features. This data.frame exists if finalCV = T.

permutations

this data.frame contains the following results of permutation experiments:

  • perm_x all RI values obtained from all permutation experiments;

  • RI RI obtained for reference MCFS experiment (i.e, the experiment on the original data); p-values from Anderson-Darling normality test applied separately for each feature to the cutoffPermutations RI set;

  • t_test_p p-values from Student-t test applied separately for each feature to the cutoffPermutations RI vs. reference RI. This data.frame exists if parameter cutoffPermutations > 0.

jrip

classification rules produced by ripper algorithm and related cross validation result obtained for top features.

params

all settings used by MCFS-ID.

exec_time

execution time of MCFS-ID.

References

M. Draminski, J. Koronacki (2018),"rmcfs: An R Package for Monte Carlo Feature Selection and Interdependency Discovery", Journal of Statistical Software, vol 85(12), 1-28, doi:10.18637/jss.v085.i12

Examples

  ## Not run: ###dontrunbegin

  ####################################
  ######### Artificial data ##########
  ####################################
  # Set VM size for Java
  options(java.parameters = "-Xmx8g")
  library(rmcfs)
  
  # create input data and review it
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 3, featureFreq = 50,
                  buildID = TRUE, finalCV = FALSE, finalRuleset = FALSE, 
                 threadsNumber = 2)

  # Print basic information about mcfs result
  print(result)
  
  # Review cutoff values for all methods
  print(result$cutoff)
  
  # Review cutoff value used in plots
  print(result$cutoff_value)
  
  # Plot & print out distances between subsequent projections. 
  # These are convergence MCFS-ID statistics.
  plot(result, type = "distances")
  print(result$distances)
  
  # Plot & print out 50 most important features and show max RI values from 
  # permutation experiment.
  plot(result, type = "ri", size = 50)
  print(head(result$RI, 50))
  
  # Plot & print out 50 strongest feature interdependencies.
  plot(result, type = "id", size = 50)
  print(head(result$ID, 50))
  
  # Plot features ordered by RI. Parameter 'size' is the number of 
  # top features in the chart. By default it is set on cutoff_value + 10
  plot(result, type = "features", cex = 1)

  # Here we set 'size' at fixed value 10.
  plot(result, type = "features", size = 10)
  
  # Plot cv classification result obtained on top features.
  # In the middle of x axis red label denotes cutoff_value.
  # plot(result, type = "cv", cv_measure = "wacc", cex = 0.8)
  
  # Plot & print out confusion matrix. This matrix is the result of 
  # all classifications performed by all decision trees on all s*t datasets.
  plot(result, type = "cmatrix")
  
  # build interdependencies graph (all default parameters).
  gid <- build.idgraph(result)
  plot(gid, label_dist = 1)
  
  # build interdependencies graph for top 6 features 
  # and top 12 interdependencies and plot all nodes
  gid <- build.idgraph(result, size = 6, size_ID = 12, orphan_nodes = TRUE)
  plot(gid, label_dist = 1)

  # Export graph to graphML (XML structure)
  path <- tempdir()
  igraph::write_graph(gid, file = file.path(path, "artificial.graphml"), 
              format = "graphml", prefixAttr = FALSE)
  
  # Export and import results to/from csv files
  export.result(result, path = path, label = "artificial")
  result <- import.result(path = path, label = "artificial")

  # Find out how many trees with the given attribute has been built (and nodes based the 
  # attribute in total). Notice that result$RI$projections keeps the number of subsets where
  # the feature was randomly picked. The value: result$RI$projections*result$params$mcfs.splits
  # is the total number of trees for a given attribute that could be built based on the attribute.
  # This normalization takes into the consideration not the full number of st trees
  # but only the fraction that is trained on datasets with the attribute.

  
  result$RI$classifiers*(result$RI$projections*result$params$mcfs.splits)	
  result$RI$nodes*(result$RI$projections*result$params$mcfs.splits)


  ####################################
  ########## Alizadeh data ###########
  ####################################
  
  # Load Alizadeh dataset.
  # A 4026 x 62 gene expression data matrix of log-ratio values. The last column contains 
  # the annotations of the 62 samples with respect to the cancer types C, D, F.
  # The data are from the lymphoma/leukemia study of A. Alizadeh et al., Nature 403:503-511 (2000), 
  # http://llmpp.nih.gov/lymphoma/index.shtml
  
  alizadeh <- read.csv(file="http://home.ipipan.waw.pl/m.draminski/files/data/alizadeh.csv", 
                        stringsAsFactors = FALSE)
  showme(alizadeh)
  
  # Fix data types and data values - replace characters such as "," " " "/" etc. 
  # from values and column names and fix data types
  # This function may help if mcfs has any problems with input data
  alizadeh <- fix.data(alizadeh)
  
  # Run MCFS-ID procedure on default parameters. 
  # For larger real data (thousands of features) default 'auto' settings are the best.
  # This example may take 10-20 minutes but this one is a real dataset with 4026 features.
  # Set up more threads according to your CPU cores number.
  result <- mcfs(class~., alizadeh, featureFreq = 100, cutoffPermutations = 10, threadsNumber = 8)
  
  # Print basic information about mcfs result.
  print(result)
  
  # Plot & print out distances between subsequent projections. 
  plot(result, type="distances")
  
  # Show RI values for top 500 features and max RI values from permutation experiment.
  plot(result, type = "ri", size = 500)
  
  # Plot heatmap on top features, only numeric features are presented
  plot(result, type = "heatmap", size = 20, heatmap_norm = 'norm', heatmap_fun = 'median')
  
  # Plot cv classification result obtained on top features.
  # In the middle of x axis red label denotes cutoff_value.
  plot(result, type = "cv", cv_measure = "wacc", cex = 0.8)
  
  # build interdependencies graph.
  gid <- build.idgraph(result, size = 20)
  plot.idgraph(gid, label_dist = 0.3)
  
  
## End(Not run)###dontrunend

Plots interdependencies graph

Description

Invokes plot.igraph with predefined parameters to visualize interdependencies graph (ID-Graph). Standard plot function with custom parameters may be used instead of this one.

Usage

  ## S3 method for class 'idgraph'
plot(x, 
          label_dist = 0.5, 
          color = 'darkred',
          cex = 1, ...)

Arguments

x

idgraph/igraph S3 object representing feature interdependencies. This object is produced by build.idgraph function.

label_dist

space between the node's label and the corresponding node in the plot.

color

it defines color of the graph nodes.

cex

size of fonts.

...

additional plotting parameters.

Examples

  ## Not run: ###dontrunbegin
  
  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 50,
                  finalCV = FALSE, finalRuleset = FALSE, threadsNumber = 2)
  
  # build interdependencies graph for top 6 features 
  # and top 12 interdependencies and plot all nodes
  gid <- build.idgraph(result, size = 6, size_ID = 12, orphan_nodes = TRUE)
  plot(gid, label_dist = 1)
  
  
## End(Not run)###dontrunend

Plots various MCFS result components

Description

Plots various aspects of the MCFS-ID result.

Usage

  ## S3 method for class 'mcfs'
plot(x, type = c("features", "ri", "id", "distances", "cv", "cmatrix", "heatmap"), 
        size = NA, 
        ri_permutations = c("max", "all", "sorted", "none"),
        diff_bars = TRUE,
        features_margin = 10,
        cv_measure = c("wacc", "acc", "pearson", "MAE", "RMSE", "SMAPE"),
        heatmap_norm = c('none', 'norm', 'scale'),
        heatmap_fun = c('median', 'mean'),
        color = c('darkred'),
        gg = TRUE,
        cex = 1, ...)

Arguments

x

'mcfs' S3 object - result of the MCFS-ID experiment returned by mcfs function.

type

  • features plots top features set along with their RI. It is a horizontal barplot that shows important features in red color and unimportant in grey.

  • ri plots top features set with their RIs as well as max RI obtained from permutation experiments. Red color denotes important features.

  • id plots top ID values obtained from the MCFS-ID.

  • distances plots distances (convergence diagnostics of the algorithm) between subsequent feature rankings obtained during the MCFS-ID experiment.

  • cv plots cross validation results based on top features.

  • cmatrix plots the confusion matrix obtained on all s \cdot t trees.

  • heatmap plots heatmap results based on top features. Only numeric features can be presented on the heatmap.

size

number of features to plot.

ri_permutations

if type = "ri" and ri_permutations = "max", then it additionally shows horizontal lines that correspond to max RI values obtained from each single permutation experiment.

diff_bars

if type = "ri" or type = "id" and diff_bars = T, then it shows difference values for RI or ID values.

features_margin

if type = "features", then it determines the size of the left margin of the plot.

cv_measure

if type = "cv", then it determines the type of accuracy shown in the plot: weighted or unweighted accuracy ("wacc" or "acc"). If target attribute is numeric it is possible to review one of the following prediction quality measures: ("pearson", "MAE", "RMSE", "SMAPE")

heatmap_norm

if type = "heatmap", then it defines type of input data normalization 'none' - without any normalization, 'norm' - normalization within range [-1,1], 'scale' - standardization/centering by mean and stdev.

heatmap_fun

if type = "heatmap", then it determines calculation 'mean' or 'median' within the class to be shown as heatmap color intensity.

color

it defines main color of the following type of plots: 'ri', 'id', 'heatmap', 'features' and 'cmatrix'.

gg

if gg = TRUE use ggplot2.

cex

size of fonts.

...

additional plotting parameters.

Examples

  ## Not run: ###dontrunbegin

  # Create input data.
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure.
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 10,
                  finalCV = FALSE, finalRuleset = TRUE, threadsNumber = 2)

  # Plot & print out distances between subsequent projections. 
  # These are convergence MCFS-ID statistics.
  plot(result, type = "distances")
  print(result$distances)
  
  # Plot & print out 50 most important features and show max RI values from 
  # permutation experiment.
  plot(result, type = "ri", size = 50)
  print(head(result$RI, 50))
  
  # Plot & print out 50 strongest feature interdependencies.
  plot(result, type = "id", size = 50)
  print(head(result$ID, 50))
  
  # Plot features ordered by RI. Parameter 'size' is the number of 
  # top features in the chart. By default it is set on cutoff_value + 10
  plot(result, type = "features", cex = 1)

  # Here we set 'size' at fixed value 10.
  plot(result, type = "features", size = 10)
  
  # Plot cv classification result obtained on top features.
  # In the middle of x axis red label denotes cutoff_value.
  # plot(result, type = "cv", measure = "wacc", cex = 0.8)
  
  # Plot & print out confusion matrix. This matrix is the result of 
  # all classifications performed by all decision trees on all s*t datasets.
  plot(result, type = "cmatrix")
  
  
## End(Not run)###dontrunend

Prints mcfs result

Description

Prints basic information about the MCFS-ID result: top features, cutoff values, confusion matrix obtained for s \cdot t trees and classification rules obtained by Ripper (jrip) algorithm.

Usage

  ## S3 method for class 'mcfs'
print(x, ...)

Arguments

x

'mcfs' object - result of the MCFS-ID experiment returned by mcfs function.

...

additional printing parameters.

Examples

  ## Not run: ###dontrunbegin

  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 10,
                  finalCV = FALSE, finalRuleset = TRUE, threadsNumber = 2)

  # Print basic information about mcfs result.
  print(result)
  
  
## End(Not run)###dontrunend

Filters input data

Description

Selects columns from input data based on the highest RIs of attributes.

Usage

prune.data(x, mcfs_result, size = NA)

Arguments

x

input data.frame.

mcfs_result

result from mcfs function.

size

number of top features to select from input data. If size = NA, then it is defined by mcfs_result$cutoff_value parameter.

Value

data.frame with selected columns.

Examples

  ## Not run: ###dontrunbegin

  # create input data
  adata <- artificial.data(rnd_features = 10)
  showme(adata)
  
  # Parametrize and run MCFS-ID procedure
  result <- mcfs(class~., adata, cutoffPermutations = 0, featureFreq = 10,
                  finalCV = FALSE, finalRuleset = FALSE, threadsNumber = 2)

  head(prune.data(adata, result, size = result$cutoff_value))

  
## End(Not run)###dontrunend

Reads data from ADH

Description

Imports data from ADH format. This format is based on two files: 'adh' that contains ADX header and 'csv' that contains the data.

Usage

read.adh(file = "")

Arguments

file

exported filename

Examples

  ## Not run: ###dontrunbegin

  # create artificial data
  adata <- artificial.data(rnd_features = 1000)

  write.adh(adata, file = file.path(tempdir(), "adata.adh"), target = "class")
  adata <- read.adh(file = file.path(tempdir(), "adata.adh"))
  
  
## End(Not run)###dontrunend

Reads data from ADX

Description

Imports data from ADX format.

Usage

read.adx(file = "")

Arguments

file

exported filename

Examples

  ## Not run: ###dontrunbegin

  # create artificial data
  adata <- artificial.data(rnd_features = 1000)
  
  write.adx(adata, file = file.path(tempdir(), "adata.adx"), target = "class")
  adata <- read.adx(file = file.path(tempdir(), "adata.adx"))
  
  
## End(Not run)###dontrunend

Basic data information

Description

Prints basic information about the data.frame.

Usage

showme(x, size = 10, show = c("tiles", "head", "tail", "none"))

Arguments

x

input data frame.

size

number of rows/columns to be printed.

show

parameters that controls print content.

  • tiles - shows top left and bottom right cells (size of both subsets is controlled by size parameter)

  • head - shows top size rows

  • tail - shows bottom size rows

  • none - does not show the content

Examples


  # create artificial data
  adata <- artificial.data(rnd_features = 1000)
  showme(adata)

Writes data to ADH

Description

Exports data into ADH format. This format is based on two files: 'adh' that contains ADX header and 'csv' that contains the data.

Usage

write.adh(x, file = "", target = NA, chunk_size = 100000, zip = FALSE)

Arguments

x

data frame with data

file

exported filename

target

sets target attribute in ADH format. Default value is NA what refers to the last column.

chunk_size

defines size of chunk (number of cells) that are processed and exported. The bigger the value, the function is faster for small data and slower for big data.

zip

whether to create zip archive.

Examples

  ## Not run: ###dontrunbegin

  # create artificial data
  adata <- artificial.data(rnd_features = 1000)
  
  #Fix input data to be consistent with ARFF and ADX formats. 
  #It is not necessary but for some data can help to export in proper format.
  adata <- fix.data(adata)
  write.adh(adata, file = file.path(tempdir(), "adata.adh"), target = "class")
  
  
## End(Not run)###dontrunend

Writes data to ADX

Description

Exports data into ADX format.

Usage

write.adx(x, file = "", target = NA, chunk_size = 100000, zip = FALSE)

Arguments

x

data frame with data

file

exported filename

target

sets target attribute in ADX format. Default value is NA what refers to the last column.

chunk_size

defines size of chunk (number of cells) that are processed and exported. The bigger the value, the function is faster for small data and slower for big data.

zip

whether to create zip archive.

Examples

  ## Not run: ###dontrunbegin

  # create artificial data
  adata <- artificial.data(rnd_features = 1000)
  
  #Fix input data to be consistent with ARFF and ADX formats. 
  #It is not necessary but for some data can help to export in proper format.
  adata <- fix.data(adata)
  write.adx(adata, file = file.path(tempdir(), "adata.adx"), target = "class")
  
  
## End(Not run)###dontrunend

Writes data to ARFF

Description

Exports data into ARFF format. This format is used by Weka Data Mining software https://git.cms.waikato.ac.nz/weka/weka.

Usage

write.arff(x, file = "", target = NA, chunk_size=100000, zip = FALSE)

Arguments

x

data frame with data

file

exported filename

target

sets target attribute in ARFF format. Default value is NA what refers to the last column.

chunk_size

it defines size of chunk (number of cells) that are processed and exported. The bigger the value, the function is faster for small data and slower for big data.

zip

whether to create zip archive.

Examples

  ## Not run: ###dontrunbegin

  # create artificial data
  adata <- artificial.data(rnd_features = 1000)
  
  #Fix input data to be consistent with ARFF and ADX formats. 
  #It is not necessary but for some data can help to export in proper format.
  adata <- fix.data(adata)
  write.arff(adata, file = file.path(tempdir(), "adata.arff"), target = "class")
  
  
## End(Not run)###dontrunend