Study and Proposal for an Improved Method of Semi-Supervised Clustering

Please enable javascript in your browser.

Abdollahi Alibeik, Mohammad | 2010

492 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 40374 (02)
University: Sharif University of Technology
Department: Mathematical Sciences
Advisor(s): Mahdavi Amiri, Nezameddin; Abolhassani, Hassan
Abstract:
Nowadays, clustering is one of the most common data mining tasks used for data categorization upon their similarities and analysis of these data groups in both industry and academia is of interest. Clustering does not need any supervision. The theme of this thesis is using some prior knowledge to improve clustering algorithms. The prior knowledge is in the form of some constraints determined by supervision to allowing preventing some combination data to be in one cluster or allocate some data in the same cluster. Here, prior knowledge in the form of constraints is used to modify a heuristic optimization algorithm (harmony search) and a novel "semi-supervised harmony clustering" is proposed. Finally, experimental results confirm the effectiveness of proposed algorithm.
Keywords:
Clustering ; Optimization ; Harmony Search Algorithm ; Semi-Supervised Clustering

No TOC