Clustering Based on Genetic Algorithms

Abstract

Clustering is an important abstraction process and it plays a vital role in both pattern recognition and data mining. Partitional algorithms are frequently used for clustering large data sets. K-means algorithm is the most popular partitional clustering algorithm; its fuzzy, rough, probabilistic and neural network are also popular. However, a major problem with the K-means algorithm and its variants is that they may not reach the globally optimal solution of the associated clustering problem. Several stochastic search techniques have been suggested in the past to address this problem. Genetic algorithms (GAs) are attractive to solve the partitional clustering problem. However, conventional GA based solutions may not scale well. A recent proposal in the literature is to use a Quad-tree based algorithm for scaling up the clustering algorithm. Unfortunately this solution does not scale up to handle large dimensional data sets. In this chapter, we explain the GA based clustering approaches and propose an efficient scheme for clustering high-dimensional largescale data sets using GAs based on the well-known CF-Tree data structure. We also discuss the notion of multi-objective clustering.