Clustering

Setup environment

#toInstall <- c("cluster", "fpc", "mclust")
#install.packages(toInstall, dependencies=TRUE)
#set working directory
#setwd("surfdrive/publications 2023/review paper UPM")
## Binary Matrix of Objectives 

Binary Matrix

A Binary Matrix has been prepared with 140 observations and 11 objectives, with presence/absence values 1/0.

mydata <- read.csv("datasetforR.csv", header = TRUE)

Using the method ‘binary’ the distance of the Matrix will be calculared

d_binary <- dist(mydata, method="binary") # First, construct a distance matrix.

Hierarchical clustering

With the distance matrix the clustering is done using the complete method for balanced and tighers clusters

fit <- hclust(d_binary, method="complete")  

Dendogram Plots

We first need to select a number of clusters to explore different dedongrams

clusters <- 8

Then we can draw the dendogram

par(mfrow=c(1,1)) # Set the plot to 1 row, 2 columns
plot(fit, cex=0.8) # cex refers to font size in the plot
rect.hclust(fit, k=clusters) # k refers to the number of clusters

Evaluation of clusters

To evaluate if the clusters are similar enough we calculate the average of Jaccard similarity index. We boostrap 1000 times.

library(fpc)
bsamples <- 1000
clus.boot <- clusterboot(mydata, 
                         B=bsamples, # Number of bootstrap resamples
                         clustermethod=hclustCBI, # for hierarchical clustering 
                         method="complete",
                         dmethod="binary",
                         cmethod="complete", # use what we used in "hclust"
                         k=clusters, 
                         count=FALSE) # Show progress on screen?

set.seed(8675309)

AvgJaccard <- clus.boot$bootmean
Instability <- clus.boot$bootbrd/bsamples
Clusters <- c(1:clusters)
Eval <- cbind(Clusters, AvgJaccard, Instability)
Eval

##      Clusters AvgJaccard Instability
## [1,]        1  0.3930995       0.854
## [2,]        2  0.4241156       0.762
## [3,]        3  0.5579577       0.461
## [4,]        4  0.6140642       0.451
## [5,]        5  0.2746869       0.985
## [6,]        6  0.1926094       0.995
## [7,]        7  0.9401069       0.017
## [8,]        8  0.3627914       0.858