Clustering technique for conceptual cluster
Brice Govin
Clustering technique for conceptual cluster
• Clustering ?
• Concept?
• An iteration to rule them all
• And in an approach bind them
• And come the hobbits and their issues
Clustering?
Clustering?
Clustering?
Clustering?
Concept
?
Concept?Conceptual cluster
Clustering concepts
An iteration to rule them allExtracted graph
A
E
IB
D
FH
G
C
D Packages
Imports
An iteration to rule them allKernel Selection
A
E
IB
D
FH
G
C
An iteration to rule them allNavigating the graph
A
E
IB
D
FH
G
C
An iteration to rule them allNavigating the graph
A
E
IB
D
FH
G
C
An iteration to rule them allNavigating the graph
A
E
IB
D
FH
G
C
An iteration to rule them allNavigating the graph
A
E
IB
D
FH
G
C
G
An iteration to rule them allDifference between traditional methods
and ours
A
E
IB
D
FH
G
C
dist(A,F)
similarity(A,H)
And in an approach bind them
And in an approach bind themWhy iterative?
Because hierarchy of clusters
Cluster 1
Cluster 1.3
Cluster 2.1
Cluster 3.1
Cluster 3.2
Cluster 2
Cluster 3
Cluster 1.2
Cluster 1.1
Cluster 2.2
And in an approach bind themHow it works?
First instance, packages clustering
Packages
Subprograms
And in an approach bind themHow it works?
First instance, packages clustering
Packages
Cluster 1
Cluster 2
Cluster 3
And in an approach bind themHow it works?
2nd instance, subprograms clustering
Packages
Subprograms
And in an approach bind themHow it works?
2nd instance, subprograms clustering
Subprograms
Cluster 1.3
Cluster 2.1
Cluster 3.1
Cluster 3.2
Cluster 1.2
Cluster 1.1
Cluster 2.2
And come the hobbits and their issuesCyclic dependencies
A
E
IB
D
FH
G
C
And come the hobbits and their issuesCyclic dependencies
Keep track of navigated nodes
A
E
IB
D
FH
G
C
And come the hobbits and their issuesMultiple allocation
A
E
IB
D
FH
G
C
Talk about numbers
1st Instance: packages clustering
2nd Instance: subprograms clustering
Precision: 98%
Recall: 88%
Precision: ~60%
Recall: ~50%%
What next?