A combination of genetic algorithm and simulated evolution techniques for clustering
Jay N. Bhuyan · 1995
Clustering is a process by which an object space is partitioned into different classes such that some optimization criteria are satisfied.The exact solution to wide variety of problems in clustering takes an exponential amount of time.Therefore researchers have developed several heuristics to solve the clustering problems.Some of these heuristics use stochastic techniques such as genetic algorithm, simulated annealing and simulated evolution to obtain near-optimal solutions with a reduced time complexity.In this paper, we present an algorithm involving a combination of genetic algorithm (GA) and simulated evolution.Each string in the GA's population is a solution state and consists of a number of clusters.The global clustering procedure is based on principles of evolution, while within each population the individuals are generated through a combination of genetic crossover operator and a further constructive step.Experimental evaluation shows that the proposed strategy produces better results than that of the best greeti heuristics known to us and has a potential for further improvement.