dissplot {seriation}  R Documentation 
Visualizes a dissimilarity matrix using seriation and matrix shading using the method developed by Hahsler and Hornik (2011). Entries with lower dissimilarities (higher similarity) are plotted darker. Dissimilarity plots can be used to uncover hidden structure in the data and judge cluster quality.
# gridbased dissimilarity plot dissplot(x, labels = NULL, method = "Spectral", control = NULL, lower_tri = TRUE, upper_tri = "average", cluster_labels = TRUE, cluster_lines = TRUE, reverse_columns = FALSE, options = NULL, ...) # ggplot2based dissimilarity plot ggdissplot(x, labels = NULL, method = "Spectral", control = NULL, lower_tri = TRUE, upper_tri = "average", cluster_labels = TRUE, cluster_lines = TRUE, reverse_columns = FALSE, ...)
x 
an object of class 
labels 

method 
A single character string indicating the seriation method used
to reorder the clusters (inter cluster seriation) as well as the objects within each
cluster (intra cluster seriation).
If different algorithms for inter and intra cluster seriation are
required, Set method to A third list element (named 
control 
a list of control options passed on to the seriation
algorithm.
In case of two different seriation algorithms, 
upper_tri, lower_tri 
a logical indicating whether to show the upper or lower triangle of the distance matrix. The string "average" can also be used to display within and between cluster averages. 
cluster_labels 
a logical indicating whether to display cluster labels in the plot. 
cluster_lines 
a logical indicating whether to draw lines to separate clusters. 
reverse_columns 
a logical indicating if the clusters are displayed
on the diagonal from
northwest to southeast ( 
options 
a list with options for plotting the matrix (

... 

The plot can also be used to visualize cluster quality (see Ling 1973). Objects belonging to the same cluster are displayed in consecutive order. The placement of clusters and the within cluster order is obtained by a seriation algorithm which tries to place large similarities/small dissimilarities close to the diagonal. Compact clusters are visible as dark squares (low dissimilarity) on the diagonal of the plot. Additionally, a Silhouette plot (Rousseeuw 1987) is added. This visualization is similar to CLUSION (see Strehl and Ghosh 2002), however, allows for using arbitrary seriating algorithms.
Note: Since pimage
uses grid, it should not be mixed with base R primitive plotting
functions, but the appropriate functions in gridpackage
.
dissplot()
returns an invisible object of class cluster_proximity_matrix
with the following
elements:
order 

cluster_order 

method 
vector of character strings indicating the seriation methods
used for plotting 
k 

description 
a 
This object can be used for plotting via
plot(x, options = NULL, ...)
, where x
is the
object and options
contains a list with plotting options (see above).
ggdissplot
returns a ggplot2 object representing the plot.
Michael Hahsler
Hahsler, M. and Hornik, K. (2011): Dissimilarity plots: A visual exploration tool for partitional clustering. Journal of Computational and Graphical Statistics, 10(2):335–354. doi: 10.1198/jcgs.2010.09139
Ling, R.F. (1973): A computer generated aid for cluster analysis. Communications of the ACM, 16(6), 355–361. doi: 10.1145/362248.362263
Rousseeuw, P.J. (1987): Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20(1), 53–65. doi: 10.1016/03770427(87)901257
Strehl, A. and Ghosh, J. (2003): Relationshipbased clustering and visualization for highdimensional data mining. INFORMS Journal on Computing, 15(2), 208–230. doi: 10.1287/ijoc.15.2.208.14448
gridpackage
,
dist
,
seriate
,
pimage
and
hmap
.
data("iris") # shuffle rows x_iris < iris[sample(seq(nrow(iris))), 5] d < dist(x_iris) # Plot original matrix dissplot(d, method = NA) # Plot reordered matrix using the nearest insertion algorithm (from tsp) dissplot(d, method = "TSP", main = "Seriation (TSP)") # Cluster iris with kmeans and 3 clusters and reorder the dissimality matrix l < kmeans(x_iris, centers = 3)$cluster dissplot(d, labels = l, main = "kmeans") # show only distances as lower triangle dissplot(d, labels = l, main = "kmeans", lower_tri = TRUE, upper_tri = FALSE) # Use a grid layout to place several plots on a page library("grid") grid.newpage() pushViewport(viewport(layout=grid.layout(nrow = 2, ncol = 2), gp = gpar(fontsize = 8))) pushViewport(viewport(layout.pos.row = 1, layout.pos.col = 1)) # Visualize the clustering (using Spectral between clusters and MDS within) res < dissplot(d, l, method = list(inter = "Spectral", intra = "MDS"), main = "KMeans + Seriation", newpage = FALSE) popViewport() pushViewport(viewport(layout.pos.row = 1, layout.pos.col = 2)) # More visualization options. Note that we reuse the reordered object res! # color: use 10 shades redblue, biased towards small distances plot(res, main = "KMeans + Seriation (redblue + biased)", col= bluered(10, bias = .5), newpage = FALSE) popViewport() pushViewport(viewport(layout.pos.row = 2, layout.pos.col = 1)) # Threshold (using zlim) and cubic scale to highlight differences plot(res, main = "KMeans + Seriation (cubic + threshold)", zlim = c(0, 2), col = grays(100, power = 3), newpage = FALSE) popViewport() pushViewport(viewport(layout.pos.row = 2, layout.pos.col = 2)) # Use gray scale with logistic transformation plot(res, main = "KMeans + Seriation (logistic scale)", col = gray( plogis(seq(max(res$x_reordered), min(res$x_reordered), length.out = 100), location = 2, scale = 1/2, log = FALSE) ), newpage = FALSE) popViewport(2) # The reordered_cluster_dissimilarity_matrix object res names(res) # ggplotbased dissplot if (require("ggplot2")) { library("ggplot2") # Plot original matrix ggdissplot(d, method = NA) # Plot seriated matrix ggdissplot(d, method = "TSP") + labs(title = "Seriation (TSP)") # Cluster iris with kmeans and 3 clusters l < kmeans(x_iris, centers = 3)$cluster ggdissplot(d, labels = l) + labs(title = "Kmeans + Seriation") # show only lower triangle ggdissplot(d, labels = l, lower_tri = TRUE, upper_tri = FALSE) + labs(title = "Kmeans + Seriation") # No lines or cluster labels and add a label for the color key (fill) ggdissplot(d, labels = l, cluster_lines = FALSE, cluster_labels = FALSE) + labs(title = "Kmeans + Seriation", fill = "Distances\n(Euclidean)") # Diverging color palette with manual set midpoint and different seriation methods ggdissplot(d, l, method = list(inter = "Spectral", intra = "MDS")) + labs(title = "KMeans + Seriation", subtitle = "redblue + biased color scale") + scale_fill_gradient2(low = "darkred", high = "darkblue", midpoint = median(d)) # Use manipulate scale using package scales library("scales") # Threshold (using limit and na.value) and cubic scale to highlight differences cubic_dist_trans < trans_new( name = "cubic", # note that we have to do the inverse transformation for distances trans = function(x) x^(1/3), inverse = function(x) x^3 ) ggdissplot(d, l, method = list(inter = "Spectral", intra = "MDS")) + labs(title = "KMeans + Seriation", subtitle = "cubic + biased color scale") + scale_fill_gradient(low = "black", high = "white", limit = c(0,2), na.value = "white", trans = cubic_dist_trans) # Use gray scale with logistic transformation logis_2_.5_dist_trans < trans_new( name = "Logistic transform (location, scale)", # note that we have to do the inverse transformation for distances trans = function(x) plogis(x, location = 2, scale = .5, log = FALSE), inverse = function(x) qlogis(x, location = 2, scale = .5, log = FALSE), ) ggdissplot(d, l, method = list(inter = "Spectral", intra = "MDS")) + labs(title = "KMeans + Seriation", subtitle = "logistic color scale") + scale_fill_gradient(low = "black", high = "white", trans = logis_2_.5_dist_trans, breaks = c(0, 1, 2, 3, 4)) }