I am working on software clustering for software module architecture recovery. When convert source code of software system in to facts, contain entities/objects and binary features. It normally contains utilities. Utilities are the entities which have all or not at all or no common features? What is the better way (and why) to detect the utilities using cluster analysis like Complete Linkage, Single linkage?
Hi Rashid,
utilitiies in software clustering are also known as omnipresent nodes. Look up the literature using this keyword. They are often identified by counting their relative incoming edges. During clustering you could weigh a node depending upon how often it is used (e.g., using Shannon Information content). Single linkage is rarely a good idea to use in software clustering because you want to get overall highly cohesive groupings.
Best regards,
Rainer
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software