ChemmineR (version 2.24.2)

jarvisPatrick: Jarvis-Patrick Clustering

Description

Function to perform Jarvis-Patrick clustering. The algorithm requires a nearest neighbor table, which consists of neighbors for each item in the dataset. This information is then used to join items into clusters with the following requirements: (a) they are contained in each other's neighbor list (b) they share at least 'k' nearest neighbors The nearest neighbor table can be computed with nearestNeighbors. For standard Jarvis-Patrick clustering, this function takes the number of neighbors to keep for each item. It also has the option of passing a cutoff similarity value instead of the number of neighbors. In this mode, all neighbors which meet the cutoff criteria will be included in the table. This is a setting that is not part of the original Jarvis-Patrick algorithm. It allows to generate tighter clusters and to minimize some limitations of this method, such as joining completely unrelated items when clustering small data sets. Other extensions, such as the linkage parameter, can also help improve the clustering quality.

Usage

jarvisPatrick(nnm,  k, mode="a1a2b", linkage="single")

Arguments

nnm

A nearest neighbor table, as produced by nearestNeighbors.

k

Minimum number of nearest neighbors two rows (items) in the nearest neighbor table need to have in common to join them into the same cluster.

mode

If mode = "a1a2b" (default), the clustering is run with both requirements (a) and (b); if mode = "a1b" then (a) is relaxed to a unidirectional requirement; and if mode = "b" then only requirement (b) is used. The size of the clusters generated by the different methods increases in this order: "a1a2b" < "a1b" < "b". The run time of method "a1a2b" follows a close to linear relationship, while it is nearly quadratic for the much more exhaustive method "b". Only methods "a1a2b" and "a1b" are suitable for clustering very large data sets (e.g. >50,000 items) in a reasonable amount of time.

linkage

Can be one of "single", "average", or "complete", for single linkage, average linkage and complete linkage merge requirements, respectively. In the context of Jarvis-Patrick, average linkage means that at least half of the pairs between the clusters under consideration must pass the merge requirement. Similarly, for complete linkage, all pairs must pass the merge requirement. Single linkage is the normal case for Jarvis-Patrick and just means that at least one pair must meet the requirement.

Value

Depending on the setting under the type argument, the function returns the clustering result in a named vector or a nearest neighbor table as matrix.

Details

...

References

Jarvis RA, Patrick EA (1973) Clustering Using a Similarity Measure Based on Shared Near Neighbors. IEEE Transactions on Computers, C22, 1025-1034. URLs: http://davide.eynard.it/teaching/2012_PAMI/JP.pdf, http://www.btluke.com/jpclust.html, http://www.daylight.com/dayhtml/doc/cluster/index.pdf

See Also

Functions: cmp.cluster trimNeighbors nearestNeighbors

Examples

## Load/create sample APset and FPset 
data(apset)
fpset <- desc2fp(apset)

## Standard Jarvis-Patrick clustering on APset/FPset objects
jarvisPatrick(nearestNeighbors(apset,numNbrs=6), k=5, mode="a1a2b")
jarvisPatrick(nearestNeighbors(fpset,numNbrs=6), k=5, mode="a1a2b")

## Jarvis-Patrick clustering only with requirement (b) 
jarvisPatrick(nearestNeighbors(fpset,numNbrs=6), k=5, mode="b")

## Modified Jarvis-Patrick clustering with minimum similarity 'cutoff' 
## value (here Tanimoto coefficient)
jarvisPatrick(nearestNeighbors(fpset,cutoff=0.6, method="Tanimoto"), k=2 )

## Output nearest neighbor table (matrix)
nnm <- nearestNeighbors(fpset,numNbrs=6)

## Perform clustering on precomputed nearest neighbor table
jarvisPatrick(nnm, k=5)