Identifying candidate objects using hierarchical clustering analysis

Clustering analysis has rarely been studied as a technique for object identification methods, although it has been broadly employed in data classification in a wide range of research areas. In this paper, we propose a review of clustering analysis methods and a scheme for applying hierarchical clustering analysis to facilitate identification of candidate objects in procedural source code. The study shows that clustering analysis is able to correctly group functions into meaningful clusters even though functions are written in an interleaved order. Clustering analysis can work well with the modular case and the tangled case without any additional support.