tree_split {PPCI}

Split a Leaf in a Hierarchical Clustering Model

Description

Adds an additional binary partition to an existing hierarchical clustering model produced by one of mcdc, mddc and ncutdc.

Usage

tree_split(sol, node, ...)

Arguments

sol

a clustering solution arising from one of the functions mcdc, mddc and ncutdc.

node

the node to be further partitioned. can be either an integer specifying the node number in sol$nodes or a vector of length two specifying c(depth, position at depth) of the node.

...

any modifications to parameters used in optimisation. these should have the same names and types as the corresponding arguments for the method used to construct sol.

Value

a list with the same components as sol. the $args field will reflect any changes included in ... above.

Examples

## load the optidigits dataset
data(optidigits)

## cluster using minimum normalised cut hyperplanes,
## assuming no domain knowledge begin with 8 clusters
sol <- ncutdc(optidigits$x, 8)

## visualise solution
plot(sol)

## node 13 shows evidence of multiple clusters. Inspect this node more closely
plot(sol, node = 13)

## split node 13
sol_new <- tree_split(sol, 13)

## compare the solutions using external cluster validity metrics
cluster_performance(sol$cluster, optidigits$c)

cluster_performance(sol_new$cluster, optidigits$c)