Title :
A New Selective Clustering Ensemble Algorithm
Author :
Liu Limin ; Fan Xiaoping
Author_Institution :
Sch. of Inf. Sci. & Eng., Central South Univ., Changsha, China
Abstract :
Selective clustering ensemble is usually based on the reference partition to select members of the ensemble. General method of generating reference partition is to use preliminary ensemble results, yet it cannot eliminate the influence of the inferior clustering partitions and the final clustering result is not satisfactory. In order to solve this problem, the paper proposes a new selective clustering ensemble algorithm. The new algorithm includes two points :(1) selecting the best reference partition based on clustering validity evaluation, (2)putting forward the new selection strategy and the method of member´s weight. The experimental results show that the new algorithm is effective and clustering accuracy could be significantly improved.
Keywords :
pattern clustering; clustering performance; clustering validity evaluation; member weight; preliminary ensemble; reference partition; reference partition selection strategy; selective clustering ensemble algorithm; Accuracy; Algorithm design and analysis; Clustering algorithms; Educational institutions; Iris; Partitioning algorithms; Pattern recognition; member´s weight; reference partition selection; selection strategy; selective clustering ensemble;
Conference_Titel :
e-Business Engineering (ICEBE), 2012 IEEE Ninth International Conference on
Conference_Location :
Hangzhou
Print_ISBN :
978-1-4673-2601-8
DOI :
10.1109/ICEBE.2012.17