A Survey on Multi-objective based clustering techniques for solving real life problems

Abstract

Clustering is a popular data mining technique which can be applied to a given data set to identify the data objects that belong to a single class, such that data objects in different clusters are distinct while similarity exists for data objects belonging to the same cluster. Usually, clustering techniques are based on optimizing single objective function criteria, which may not be capable of performing well in many real time scenarios. Motivated by this many multi-objective based optimization techniques are discussed in this paper. Multi-objective based optimization techniques are capable of optimizing several conflicting objective functions simultaneously. Under this context, evolutionary based approach and simulated annealing based techniques are adopted in various MOO techniques and proven well in case of noise, non-spherical and high dimensional feature space. The paper further discusses various validity measures to evaluate the goodness of clustering techniques.