IBM Frequent Subgraph Miner
A data-mining tool for discovering all frequent substructure patterns from a set of labeled graphs.
Date Posted: February 27, 2004
|
|
 |
|
|
|
IBM Frequent Subgraph Miner can analyze a set of undirected graphs that do not contain self-looped vertices. The format of input file is as follows:
size(x,y)
x: graph identifier starting with 1.
y: size, which means the number of vertices in the graph.
name(x,y)
x: graph identifier starting with 1.
y: name of the graph
node(x,y,z)
x: graph identifier starting with 1.
y: vertex identifier starting with 1.
z: vertex labels
link(x,y1,y2,z)
x: graph identifier starting with 1.
y1: vertex identifier starting with 1.
y2: vertex identifier starting with 1.
z: edge label of an edge between vertices y1 and y2.
| | |
|
Add the ?out option. All discovered patterns are put out as follows when the sample data is used:
size = 4
VertexLabels = c22,c22,h3,h3
code = r,s,0,0,s,0
count = 189
It means that one of the discovered patterns has 4 vertices, 189 graphs contain this pattern, and its structure is represented by the following adjacency matrix:
c22 c22 h3 h3
c22| 0 r s 0 |
c22| r 0 0 s |
h3 | s 0 0 0 |
h3 | 0 s 0 0 |
| | |
|
In order to reduce the candidates of the frequent subgraphs, the code of an adjacency matrix is defined as follows: The code of an adjacency matrix Xk whose size is k x k is defined as
code(Xk)=x1,2x1,3x2,3x1,4...cxk-2,kxk-1,k
by using (i,j)-element xi,j of Xk.
| |
|
|
|
|
| |
