ClusterCor: Cluster correlation for cluster evaluation
In jhmadsen/ClustTools: Clustering evaluation package

Description Usage Arguments Details Value Author(s) References Examples

Function to evaluate clustering results by calculating the correlation between an incidence matrix and distance matrix. Suggested by Tan, P.-N., Steinbach, M., Karpatne, A., & Kumar, V. (2005)

1	ClusterCor(dist.obj, clusterVector, return_matrices = FALSE)

`dist.obj`	An object of class 'dist' for dataset
`clusterVector`	A vector with integers indicating which cluster observations belong to
`return_matrices`	Argument to return incidence- and distance matrix for observations

ClusterCor computes an incidence matrix for observations, given a cluster vector, by creating a n x n matrix where 1 is returned for observations in same cluster and 0 is returned for observations in different clusters. With a distance matrix as input the two matrices are vectorized and correlation is computed. A highly negative correlation indicates that observations in same cluster have small distance to each other, meaning good results with respect to minimizing intra-distance and maximizing inter-distance

`incidenceMatrix`	Matrix with 0's and 1's indicating if observations belong to same cluster or not
`distMatrix`	Matrix with distances between observations
`correlation`	The correlation coefficient. The more negative correlation, the better results are achieved with respect to clustering objectives

Jacob H. Madsen

Tan, P.-N., Steinbach, M., Karpatne, A., & Kumar, V. (2005). Introduction to Data Mining (Second edition). ISBN: 978-03-213-2136-7

## Select a dataset to standardize and cluster
X <- scale(iris[,1:4])

## Cluster the dataset with a given number of clusters
cluster.obj <- kmeans(X, 3)

## Evaluate the clustering results with 'ClusterCor'
ClusterCor(dist(X), cluster.obj$cluster)