Memory efficient and fast calculation of Rand Index. The package offers algorithm for Rand Index estimation that is specifically meant for large datasets with many small clusters making it the fastest and most memory efficient in comparison to other available R packages (see benchmarking below).
| package | version | comment |
|---|---|---|
| randex | 0.0.0.9000 | This package |
| clusteval | 0.2.1 | |
| matchFeat | 1.0 | |
| fossil | 0.4.0 | |
| ClustOfVar | 1.1 | |
| hecmulti | 2023.11.19 | Documentation in French:) |
| RFclust.SGE | 0.0.0.9000 | Does not work. Filed an issue |
Rand index estimation
Rand index measures similarity between two partitionings X and Y (a.k.a., clusterings) of a set of n elements S. It can also be seen as a probability of a random pair of elements from S to be either in a same subset/cluster in both partitions or to be in different subsets/clusters in both partitions. It calculates as following:
Rand Index = \frac {TP + TN} {TP + FP + FN + TN} = 2 \times \frac {TP + TN} {n \times (n-1)}, where:
- TP is the number of true positives, i.e., a number of pairs of elements in S that are in the same subset for both partitionins X and Y
- TN is the number of true negatives, i.e., a number of pairs of elements in S that are in different subsets for both partitionins X and
- FP is the number of false positives, i.e., a number of pairs of elements in S that are in same subset in X (assuming that X represents ground thuth parititioning) and are in different subsets in Y
- FN is the number of false negatives, i.e., a number of pairs of elements in S that are in different subsets in X and are in same subset in Y
The calculation of Rand Index implemented in randex package is slightly different from the straightforward calculation that follows the the above definition. Since it is easier to compute disagreements between partitioning (i.e., FP and FN) rather than agreements (TP + TN) we can calculate it as $Rand Index = 1 - 2 \times \frac {FP + FN} {n \times (n-1)}$. The package relies on data.table features for efficient data manipulation in memory.
References
ClustOfVar- ClustOfVar: An R Package for the Clustering of Variables | Journal of Statistical Software
matchFeat - Degras (2021). Scalable Feature Matching Across Large Data Collections. https://arxiv.org/abs/2101.02035 Scalable Feature Matching Across Large Data Collections: Journal of Computational and Graphical Statistics: Vol 32 , No 1 - Get Access Local search heuristics for multi-index assignment problems with decomposable costs: Journal of the Operational Research Society: Vol 55 , No 7 - Get Access
fossil-
@Article{,
title = {fossil: palaeoecological and palaeogeographical analysis
tools},
author = {Matthew J. Vavrek},
year = {2011},
journal = {Palaeontologia Electronica},
volume = {14},
pages = {1T},
number = {1},
note = {R package version 0.4.0},
}Dependencies
| name | version | comment |
|---|---|---|
| R | 4.2.0 | minimum R version to enable native piping |
Hard dependencies (Depends field in DESCRIPTION file)
| name | version | comment |
|---|---|---|
| data.table | fast data.frames, used as main input and output data type |
Required packages (Imports field in the DESCRIPTION file)
| name | version | comment |
|---|---|---|
| tinytest | package development (unit testing) | |
| ggplot2 | for visualizing benchmarks | |
| patchwork | for combining plots of memory and time benchmarks |
Suggested packages (Suggests field in the DESCRIPTION file)
Benchmark
The source code for benchmarking is below. All packages used in this benchmark can be fully reproduced with Guix package manager. The module with the packages definitions is below.
some results (to sto)
[2024-02-14 Wed] ## N <- 22:25 lentgh 3
Evaluating: matchFeat::Rand.index(a, b) call sample result memory elapsed (sec) [1,] “randex::randindex(a, b)” 4194304 NA 1682838448 11.607 [2,] “matchFeat::Rand.index(a, b)” 4194304 0.9998718 4334704504 8.563 $call
Guix module with R packages for Benchmark
(define-module (my packages r)
#:use-module (gnu packages)
#:use-module (gnu packages statistics)
#:use-module (guix packages)
#:use-module (guix download)
#:use-module (guix build utils)
#:use-module (guix build-system r)
#:use-module (guix build-system trivial)
#:use-module (guix licenses)
;; #:use-module (git)
;; #:use-module (guix git)
#:use-module (guix git-download)
)
;; guix import cran --recursive ClustOfVar
(define-public r-pcamixdata
(package
(name "r-pcamixdata")
(version "3.1")
(source
(origin
(method url-fetch)
(uri (cran-uri "PCAmixdata" version))
(sha256
(base32 "0flrsnbchwk06dmkg3vqykp9n4pqs265szn1r10navp8ki3rrmvh"))))
(properties `((upstream-name . "PCAmixdata")))
(build-system r-build-system)
(native-inputs (list (specification->package "r-knitr")))
(home-page "https://cran.r-project.org/package=PCAmixdata")
(synopsis "Multivariate Analysis of Mixed Data")
(description
"This package implements principal component analysis, orthogonal rotation and
multiple factor analysis for a mixture of quantitative and qualitative
variables.")
(license gpl2+)))
(define-public r-clustofvar
(package
(name "r-clustofvar")
(version "1.1")
(source
(origin
(method url-fetch)
(uri (cran-uri "ClustOfVar" version))
(sha256
(base32 "0grhkab7s58ji4cf7cxh7ahd2dxrj8aqfdf3119b40zxkxbwxcr0"))))
(properties `((upstream-name . "ClustOfVar")))
(build-system r-build-system)
(propagated-inputs (list r-pcamixdata))
(home-page "https://cran.r-project.org/package=ClustOfVar")
(synopsis "Clustering of Variables")
(description
"Cluster analysis of a set of variables. Variables can be quantitative,
qualitative or a mixture of both.")
(license gpl2+)))
;; guix import cran --recursive fossil
(define-public r-shapefiles
(package
(name "r-shapefiles")
(version "0.7.2")
(source
(origin
(method url-fetch)
(uri (cran-uri "shapefiles" version))
(sha256
(base32 "03sdcxbah05x0j6cpygx3ivkzrdlz2c0frxi30cinb05q6a41yjb"))))
(properties `((upstream-name . "shapefiles")))
(build-system r-build-system)
(propagated-inputs (list (specification->package "r-foreign")))
(home-page "https://cran.r-project.org/package=shapefiles")
(synopsis "Read and Write ESRI Shapefiles")
(description
"This package provides functions to read and write ESRI shapefiles.")
(license (list gpl2+ gpl3+))))
(define-public r-fossil
(package
(name "r-fossil")
(version "0.4.0")
(source
(origin
(method url-fetch)
(uri (cran-uri "fossil" version))
(sha256
(base32 "1hbls9m8yapnfzpv9s850ixakmnan8min1ynk7dqkbpb2px85h1p"))))
(properties `((upstream-name . "fossil")))
(build-system r-build-system)
(propagated-inputs (list
(specification->package "r-maps" )
r-shapefiles
(specification->package "r-sp" )))
(home-page "http://matthewvavrek.com/programs-and-code/fossil/")
(synopsis "Palaeoecological and Palaeogeographical Analysis Tools")
(description
"This package provides a set of analytical tools useful in analysing ecological
and geographical data sets, both ancient and modern. The package includes
functions for estimating species richness (Chao 1 and 2, ACE, ICE, Jacknife),
shared species/beta diversity, species area curves and geographic distances and
areas.")
(license gpl2+)))
;; guix import cran --style=specification --recursive matchFeat
(define-public r-matchfeat
(package
(name "r-matchfeat")
(version "1.0")
(source
(origin
(method url-fetch)
(uri (cran-uri "matchFeat" version))
(sha256
(base32 "0jh484rr71b7887igfslbg7xbr661l9c34d650xd7ajx4gfpn540"))))
(properties `((upstream-name . "matchFeat")))
(build-system r-build-system)
(propagated-inputs (list (specification->package "r-clue")
(specification->package "r-foreach")))
(home-page "https://cran.r-project.org/package=matchFeat")
(synopsis "One-to-One Feature Matching")
(description
"Statistical methods to match feature vectors between multiple datasets in a
one-to-one fashion. Given a fixed number of classes/distributions, for each
unit, exactly one vector of each class is observed without label. The goal is
to label the feature vectors using each label exactly once so to produce the
best match across datasets, e.g. by minimizing the variability within classes.
Statistical solutions based on empirical loss functions and probabilistic
modeling are provided. The Gurobi software and its R interface package are
required for one of the package functions (match.2x()) and can be obtained at
<https://www.gurobi.com/> (free academic license). For more details, refer to
Degras (2022) <doi:10.1080/10618600.2022.2074429> \"Scalable feature matching for
large data collections\" and Bandelt, Maas, and Spieksma (2004)
<doi:10.1057/palgrave.jors.2601723> \"Local search heuristics for multi-index
assignment problems with decomposable costs\".")
(license gpl2)))
;; needed for stela2502/RFclust.SGE
;; stas@air ~/dot/sys/my-guix-channel/my/packages$ guix import cran --style=specification --recursive --archive=git https://github.com/sonejilab/FastWilcoxTest >> r.scm
(define-public r-fastwilcoxtest
(let ((commit "c9ea65dcc41aa5f3403441899f7e558d2a7cbe7d")
(revision "1"))
(package
(name "r-fastwilcoxtest")
(version (git-version "0.2.0" revision commit))
(source
(origin
(method git-fetch)
(uri (git-reference
(url "https://github.com/sonejilab/FastWilcoxTest")
(commit commit)))
(file-name (git-file-name name version))
(sha256
(base32 "0fpblsarxjazmbya3lr304chhc0fwsj6xp7sa5fhi4ryqqw7zrlm"))))
(properties `((upstream-name . "FastWilcoxTest")))
(build-system r-build-system)
(inputs (list (specification->package "r-gsl")))
(propagated-inputs (list (specification->package "r-matrix")
(specification->package "r-metap")
(specification->package "r-rcpp")
(specification->package "r-rcppeigen")
(specification->package "r-rcppprogress")
(specification->package "r-reshape2")))
(home-page "https://github.com/sonejilab/FastWilcoxTest")
(synopsis
"Wilcox Ranked Sum Test Implementation using Rcpp; Tests are Applied to a Sparse Matrix")
(description
"Re-implementation the the Seurat::@code{FindMarkers}'( test.use == \"wilcox\" )
function but implementing all calculation steps in c++. Thereby the function is
more than 10 times faster than the Seurat R implementation. The c++ code was
extracted from the @code{BioQC} @code{BioConductor} package. It also contains
other fast c++ functions to interact with sparse matrices.")
(license gpl3))))
;; stas@air ~/dot/sys/my-guix-channel/my/packages$
;; guix import cran --style=specification --recursive --archive=git https://github.com/stela2502/RFclust.SGE >> r.scm
;; guix import: warning: failed to retrieve package information from https://cran.r-project.org/web/packages/FastWilcoxTest/DESCRIPTION: 404 (Not Found)
(define-public r-rfclust-sge
(let ((commit "ba586d8f0372f7ceb29b75fd3290931856ef64a8")
(revision "1"))
(package
(name "r-rfclust-sge")
(version (git-version "0.0.0.9000" revision commit))
(source
(origin
(method git-fetch)
(uri (git-reference
(url "https://github.com/stela2502/RFclust.SGE")
(commit commit)))
(file-name (git-file-name name version))
(sha256
(base32 "0hq2rdyxylm5fhbvjpq2dncpvz2m8zfdn787hxpmj10gvv5xr435"))))
(properties `((upstream-name . "RFclust.SGE")))
(build-system r-build-system)
(propagated-inputs (list (specification->package "r-cluster")
;; FastWilcoxTest
r-fastwilcoxtest
(specification->package "r-hmisc")
(specification->package "r-mass")
(specification->package "r-matrix")
(specification->package "r-ranger")
(specification->package "r-survival")))
(home-page "https://github.com/stela2502/RFclust.SGE")
(synopsis "Unsupervised clustering using random forest run on SGE")
(description
"This package uses the RF clustering method described at https://
labs.genetics.ucla.edu/horvath/RFclustering/RFclustering.htm. The function is
broken down into separate parts, that can be run on a SGE to reduce analysis
time.")
(license expat))))
;; guix import cran --style=specification --recursive --archive=git https://github.com/ramhiser/clusteval >> r.scm
(define-public r-clusteval
(let ((commit "09eae82610a13122d6bfd46480fc4a76eb3c752a")
(revision "1"))
(package
(name "r-clusteval")
(version (git-version "0.2.1" revision commit))
(source
(origin
(method git-fetch)
(uri (git-reference
(url "https://github.com/ramhiser/clusteval")
(commit commit)))
(file-name (git-file-name name version))
(sha256
(base32 "1591acinzd4mgp8sg9mn0syn1caaxdy3ys99pnpqa5yb0x423y7p"))))
(properties `((upstream-name . "clusteval")))
(build-system r-build-system)
(propagated-inputs (list (specification->package "r-ggplot2")
(specification->package "r-mvtnorm")
(specification->package "r-rcpp")))
(home-page "https://github.com/ramhiser/clusteval")
(synopsis "Evaluation of Clustering Algorithms")
(description
"This package provides a suite of tools to evaluate clustering algorithms,
clustering, and individual clusters.")
(license expat))))
;; guix import cran --style=specification --recursive --archive=git https://github.com/lbelzile/hecmulti >> r.scm
(define-public r-hecmulti
(let ((commit "7488f654ae860a1a139bc05b8c263cf7b7fb4517")
(revision "1"))
(package
(name "r-hecmulti")
(version (git-version "2023.11.19" revision commit))
(source
(origin
(method git-fetch)
(uri (git-reference
(url "https://github.com/lbelzile/hecmulti")
(commit commit)))
(file-name (git-file-name name version))
(sha256
(base32 "1v3yzfjkhhbd253p158d4rglarbv5fx9skr6ya4drx69bdbdr03v"))))
(properties `((upstream-name . "hecmulti")))
(build-system r-build-system)
(arguments
(list
#:modules '((guix build r-build-system)
(guix build minify-build-system)
(guix build utils)
(ice-9 match))
#:imported-modules `(,@%r-build-system-modules (guix build
minify-build-system))
#:phases '(modify-phases %standard-phases
(add-after 'unpack 'process-javascript
(lambda* (#:key inputs #:allow-other-keys)
(with-directory-excursion "inst/"
(for-each (match-lambda
((source . target) (minify source
#:target
target)))
'())))))))
(propagated-inputs (list (specification->package "r-ggplot2")
(specification->package "r-mass")
(specification->package "r-patchwork")))
(native-inputs (list (specification->package "esbuild")
(specification->package "r-knitr")))
(home-page "https://github.com/lbelzile/hecmulti")
(synopsis "Matériel de cours pour Analyse multidimensionnelle appliquée")
(description
"Jeux de données et fonctions pour le cours Analyse multidimensionnelle appliquée
(MATH 60602) à HEC Montréal.")
(license cc-by-sa4.0))))