Browse Prior Art Database

# Process Clustering Method

IP.com Disclosure Number: IPCOM000103863D
Original Publication Date: 1993-Feb-01
Included in the Prior Art Database: 2005-Mar-18
Document File: 2 page(s) / 42K

IBM

## Related People

Raz, T: AUTHOR [+2]

## Abstract

The invention establishes an algorithm to cluster processes based on an objective measure of their coupling. With the new method, process clusters can be systematically identified, the effort required to manage processes can be reduced, and the overall effectiveness of process management can be enhanced.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 56% of the total text.

Process Clustering Method

The invention establishes an algorithm to cluster processes
based on an objective measure of their coupling.  With the new
method, process clusters can be systematically identified, the effort
required to manage processes can be reduced, and the overall
effectiveness of process management can be enhanced.

A method for process clustering is described.  It is an
agglomerative clustering algorithm based on a coupling metric.  It
starts with each process being in a cluster by itself, and at each
iteration the two clusters that are the closest to each other among
all possible pairs of clusters are merged, with closeness measured by
the coupling metric.  This is repeated until only one cluster
remains.

Initially, a linkage matrix M is established with element Mij
representing the number of links from cluster i to cluster j.  Then,
based on the linkage matrix M, a coupling metric between cluster i
with Ni processes and cluster j with Nj processes is defined as

C(i,j) = (Mij + Mji) / (2 x Ni x Nj)

A coupling matrix C is constructed, where each element Cij =
C(i,j).  The clustering algorithm is presented in detail next.

1.  Initialize the number of processes in each cluster:
For all i, 1 <= i <= n:
Ni = 1

2.  Initialize the upper triangular portion of the coupling matrix C:
For all i and j, 1 <= i < j <= n:
Cij = C(i,j)

3.  DO n-1 times:
a.  Fi...