Projected Clustering Using Particle Swarm Optimization

Clustering methods divide the dataset into groups of similar objects, where objects in the same group are similar and objects in different groups are dissimilar. Traditional clustering techniques that find clusters in full dimensional space may fail to find clusters in high dimensional data due to various problems associated with clustering on high dimensional data. Subspace and projected clustering methods find clusters that exist in subspaces of dataset. These methods provide solutions to challenges associated with clustering on high dimensional data. Projected clustering methods output subspace clusters where one point in the dataset belongs to only one subspace cluster. Points may be assigned to multiple subspace clusters by subspace clustering methods. Projected clustering is preferable to subspace clustering when partition of points is required. Particle swarm optimization (PSO) has been proven to be effective for solving complex optimization problems. In this paper, we propose a Projected Clustering Particle Swarm Optimization (PCPSO) method to find subspace clusters that are present in the dataset. In PCPSO, Particle swarm optimization has been used to find optimal cluster centers by optimizing a subspace cluster validation index. In this paper, kmeans has been used to find neighbourhood of subspace cluster centers. The proposed method has been used to find subspace clusters that are present in some synthetic datasets. (c) 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of C3IT