Regularized KMEANS clustering, new functionalities

Implementation of hard threshold regularized KMEANS clustering (HTKmeans, function `htk_cluster()`). Tuning functions for lambda in HTKmeans and k in kNN prediction of the cluster assignment
PiotrTymoszuk · Nov 17, 2023 · 9ed2609 · 9ed2609
1 parent 3a0adc7
commit 9ed2609
Show file tree

Hide file tree

Showing 23 changed files with 2,342 additions and 299 deletions.
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,12 +1,12 @@
 Package: clustTools
 Type: Package
 Title: Tools for Dimensionality Reduction and Data Clustering
-Version: 1.2.1
+Version: 1.3.0
 Maintainer: Piotr Tymoszuk <[email protected]>
 Description: Provides a comprehensive set of R functions for dimensionality 
-    reduction (PCA, MDS and UMAP), clustering (HCl, k-means, PAM, DBSCAN, SOM)
-    and combined reduction-clustering tools. Visualization and QC toolset is 
-    provided as well.
+    reduction (PCA, MDS and UMAP), clustering (HCl, KMEANS, PAM, DBSCAN, SOM 
+    and regularized KMEANS) and combined reduction-clustering tools. 
+    Visualization and QC toolset is provided as well.
 License: GPL-3
 Encoding: UTF-8
 LazyData: true
@@ -16,6 +16,7 @@ Authors@R:
     comment = c(ORCID = "0000-0002-0398-6034"))
 Imports: 
     caret,
+    clusterHD,
     cowplot,
     coxed,
     dbscan,

diff --git a/NAMESPACE b/NAMESPACE
@@ -21,6 +21,7 @@ S3method(extract,clust_analysis)
 S3method(extract,cluster_cv)
 S3method(extract,combi_analysis)
 S3method(extract,red_analysis)
+S3method(extract,tuner)
 S3method(impact,clust_analysis)
 S3method(impact,combi_analysis)
 S3method(impact,min_analysis)
@@ -39,12 +40,14 @@ S3method(np,red_analysis)
 S3method(pbc,clust_analysis)
 S3method(pbc,combi_analysis)
 S3method(plot,clust_analysis)
+S3method(plot,clust_red)
 S3method(plot,combi_analysis)
 S3method(plot,cross_dist)
 S3method(plot,importance)
 S3method(plot,knb)
 S3method(plot,red_analysis)
 S3method(plot,sil_extra)
+S3method(plot,tuner)
 S3method(plot,umatrix_analysis)
 S3method(plot_clust_hm,clust_analysis)
 S3method(plot_clust_hm,combi_analysis)
@@ -55,22 +58,30 @@ S3method(predict,combi_analysis)
 S3method(predict,min_analysis)
 S3method(predict,red_analysis)
 S3method(predict,umatrix_analysis)
+S3method(prediter,clust_analysis)
+S3method(prediter,combi_analysis)
+S3method(prediter,min_analysis)
+S3method(prediter,umatrix_analysis)
 S3method(print,clust_analysis)
 S3method(print,combi_analysis)
 S3method(print,cross_dist)
 S3method(print,red_analysis)
+S3method(print,tuner)
 S3method(qe,clust_analysis)
 S3method(qe,combi_analysis)
 S3method(rename,clust_analysis)
 S3method(rename,combi_analysis)
 S3method(silhouette,clust_analysis)
 S3method(silhouette,combi_analysis)
+S3method(summary,clust_analysis)
 S3method(summary,cluster_cv)
+S3method(summary,combi_analysis)
 S3method(summary,cross_dist)
 S3method(summary,importance)
 S3method(summary,knb)
 S3method(summary,red_analysis)
 S3method(summary,sil_extra)
+S3method(summary,tuner)
 S3method(te,clust_analysis)
 S3method(te,combi_analysis)
 S3method(var,clust_analysis)
@@ -97,6 +108,7 @@ export(extract.red_analysis)
 export(get_clust_tendency)
 export(get_kernel_info)
 export(hcluster)
+export(htk_cluster)
 export(impact)
 export(is_clust_analysis)
 export(is_cluster_cv)
@@ -107,6 +119,7 @@ export(is_knb)
 export(is_min_analysis)
 export(is_red_analysis)
 export(is_sil_extra)
+export(is_tuner)
 export(is_umatrix_analysis)
 export(kcluster)
 export(min_max)
@@ -121,11 +134,13 @@ export(plot.importance)
 export(plot.knb)
 export(plot.red_analysis)
 export(plot.sil_extra)
+export(plot.tuner)
 export(plot_clust_hm)
 export(plot_nbclust)
 export(predict.clust_analysis)
 export(predict.combi_analysis)
 export(predict.red_analysis)
+export(prediter)
 export(qe)
 export(red_analysis)
 export(reduce_data)
@@ -136,11 +151,14 @@ export(silhouette.clust_analysis)
 export(silhouette.combi_analysis)
 export(som_cluster)
 export(som_reduce)
+export(summary.clust_analysis)
+export(summary.combi_analysis)
 export(summary.cross_dist)
 export(summary.importance)
 export(summary.knb)
 export(summary.sil_extra)
 export(te)
+export(tune_htk)
 export(var)
 export(var.clust_analysis)
 export(var.combi_analysis)

diff --git a/R/appearance.R b/R/appearance.R
@@ -1,6 +1,6 @@
 # Appearance for the S3 class objects
 
-#' Printing of a clust_analysis object.
+#' Printing of objects.
 #'
 #' @description
 #' Prints a `clust_analysis`, `combi_analysis`, `red_analysis`
@@ -33,7 +33,12 @@
 
     stopifnot(is_combi_analysis(x))
 
-    purrr::walk(x$clust_analyses, print)
+    cat(paste('Combined SOM/clusetring analysis with',
+               x$dist_method, 'distance method.'))
+
+    cat('\nCluster assignment:\n')
+
+    print(as_tibble(x$clust_assignment))
 
   }
 
@@ -61,9 +66,27 @@
     stopifnot(is_red_analysis(x))
 
     cat(paste0(toupper(x$red_fun), ' reduction analysis object.'))
+
     cat('\nComponents:\n')
+
     print(as_tibble(x$component_tbl))
 
   }
 
+#' @rdname print.clust_analysis
+#' @export
+
+  print.tuner <- function(x, ...) {
+
+    stopifnot(is_tuner(x))
+
+    cat(paste('Tuning results for',
+              paste(x$tune_params, collapse = ', ')))
+
+    cat('\nBest parameter combination:\n')
+
+    print(as_tibble(x$best_tune))
+
+  }
+
 # END ------
diff --git a/R/class_testing.R b/R/class_testing.R
@@ -5,7 +5,7 @@
 #' @description
 #' Tests if the object is an instance of the `red_analysis`, `clust_analysis`,
 #' `combi_analysis`, `importance`, `cross_dist`, `sil_extra`, `min_analysis`,
-#' `umatrix_analysis`, `cluster_cv` or `knb` class.
+#' `umatrix_analysis`, `cluster_cv`, `knb` or `tuner` class.
 #'
 #' @return a logical value.
 #'
@@ -60,4 +60,9 @@
 
   is_knb <- function(x) inherits(x, 'knb')
 
+#' @rdname is_clust_analysis
+#' @export
+
+  is_tuner <- function(x) inherits(x, 'tuner')
+
 # END ------