An evolutionary subspace clustering algorithm for high-dimensional data
Seyednaser Nourashrafeddin, Dirk V. Arnold, Evangelos Milios · 2012
We present an algorithm for generating subspace clusterings of large data sets with many attributes. An evolutionary algorithm is used to form groups of relevant attributes. Those groups are replaced by their centroids, making it possible to cluster the objects in a much lower dimensional space. Preliminary experiments with scalable synthetic data sets suggest that the algorithm generates competitive clusterings while scaling quite well.