IASTED International Joint Conference on Artificial Intelligence and Soft Computing (ASC'2004), September 1-3, 2004, Marbella, Spain
data mining; virtual reality; hybrid optimization
Visual data mining via the construction of virtual reality spaces for the representation of data and knowledge, involves the solution of optimization problems. This paper introduces a hybrid technique based on particle swarm optimization (PSO) combined with classical optimization methods. This approach is applied to very high dimensional data from microarray gene expression experiments in order to understand the structure of both raw and processed data. Experiments with data sets corresponding to Alzheimer's disease show that high quality visual representations can be obtained by combining PSO with classical optimization methods. The behaviour of some of the parameters controlling the swarm evolution is also studied.
Proceedings of the IASTED International Joint Conference on Artificial Intelligence and Soft Computing (ASC'2004).