|
In recent years, clustering ensembles have attracted much attention since they outperforms the traditional clustering methods. A lot of researches have been done both on constructing the individual partitions and on designing the consensus functions. This paper focuses on the second aspect. Namely, how to combine the multiple data partitions to get a consistent partition for a given dataset using the information obtained in the different clusterings. In this paper, we propose a new method of combining multiple partitions by using the Squared Error Adjacent Matrix (SEAM) algorithm. We conducted several experiments of the proposed method both on the synthetic and the real-world datasets and compared the method with the graph-based consensus functions, CSPA, HGPA, and MCLA. Experimental results show that the proposed method is better or comparable to the graph-based methods. |
|
Keywords:data mining; clustering; clustering ensemble |
|