k-divisive hierarchical clustering

Studenteropgave: Speciale (inkl. HD afgangsprojekt)

  • Addin Osman Mohamed Addin
4. semester, Datalogi, Kandidat (Kandidatuddannelse)
In this master thesis, a novel divisive hierarchical clustering algorithm has been proposed. The algorithm is called the $k$-divisive hierarchical clustering algorithm. The aim of this algorithm is to overcome some of the disadvantages of the well-known divisive hierarchical clustering algorithms. The proposed algorithm builds the hierarchy by splitting a cluster into tow sub-clusters. The splitting process is performed by implementing the $k$-means algorithm, and the process of splitting a cluster is stopped by using two proposed stop criteria. The splitting process is halted so as to end up with a high quality dendogram. The algorithm together with the stop criteria are implemented and tested using artificial databases as well as real world databases.
Udgivelsesdatojun. 2003
ID: 61058298