Package website: release | dev
Cluster analysis for mlr3.
mlr3cluster is an extension package for cluster analysis within the mlr3 ecosystem. It is a successor of clustering capabilities of mlr2.
Installation
Install the last release from CRAN:
install.packages("mlr3cluster")Install the development version from GitHub:
# install.packages("pak")
pak::pak("mlr-org/mlr3cluster")Feature Overview
The current version of mlr3cluster contains:
- A selection of 24 clustering learners that represent a wide variety of clusterers: partitional, hierarchical, fuzzy, etc.
- A selection of 4 performance measures
- Two built-in tasks to get started with clustering
Also, the package is integrated with mlr3viz which enables you to create great visualizations with just one line of code!
Cluster Analysis
Cluster Learners
| Key | Label | Packages |
|---|---|---|
| clust.MBatchKMeans | Mini Batch K-Means | ClusterR |
| clust.SimpleKMeans | K-Means (Weka) | RWeka |
| clust.agnes | Agglomerative Nesting | cluster |
| clust.ap | Affinity Propagation | apcluster |
| clust.bico | BICO | stream |
| clust.birch | BIRCH | stream |
| clust.cmeans | Fuzzy C-Means | e1071, clue |
| clust.cobweb | Cobweb | RWeka |
| clust.dbscan | DBSCAN | dbscan |
| clust.dbscan_fpc | DBSCAN (fpc) | fpc |
| clust.diana | Divisive Analysis | cluster |
| clust.em | Expectation-Maximization | RWeka |
| clust.fanny | Fuzzy Analysis | cluster |
| clust.featureless | Featureless Clustering Learner | |
| clust.ff | Farthest First | RWeka |
| clust.hclust | Hierarchical Clustering | stats |
| clust.hdbscan | HDBSCAN | dbscan |
| clust.kkmeans | Kernel K-Means | kernlab |
| clust.kmeans | K-Means | stats, clue |
| clust.mclust | Gaussian Mixture Model | mclust |
| clust.meanshift | Mean Shift | LPCM |
| clust.optics | OPTICS | dbscan |
| clust.pam | Partitioning Around Medoids | cluster, clue |
| clust.xmeans | X-Means | RWeka |
Cluster Measures
| Key | Label | Packages |
|---|---|---|
| clust.ch | Calinski Harabasz | fpc |
| clust.dunn | Dunn | fpc |
| clust.silhouette | Silhouette | cluster |
| clust.wss | Within Sum of Squares | fpc |
Example
library(mlr3)
library(mlr3cluster)
task = tsk("usarrests")
task
#>
#> ── <TaskClust> (50x4): US Arrests ──────────────────────────────────────────────
#> • Target:
#> • Properties: -
#> • Features (4):
#> • int (2): Assault, UrbanPop
#> • dbl (2): Murder, Rape
learner = lrn("clust.kmeans")
prediction = learner$train(task)$predict(task)
measures = msrs(c("clust.wss", "clust.silhouette"))
prediction$score(measures, task)
#> clust.wss clust.silhouette
#> 9.639903e+04 5.926554e-01More Resources
Check out the blogpost for a more detailed introduction to the package. Also, mlr3book has a section on clustering.
Future Plans
- Add more learners and measures
- Integrate the package with mlr3pipelines (work in progress)
If you have any questions, feedback or ideas, feel free to open an issue here.