Enhanced Self Organizing Map (SOM) and Particle Swarm Optimization (PSO) for Classification

Abstract

Hybrid technique for Self Organizing Map and Particle Swarm Optimization approach is commonly implemented in clustering area. In this paper, a hybrid approach that is based on Enhanced Self Organizing Map and Particle Swarm Optimization (ESOM/PSO) for classification is proposed. Enhanced Self Organization map which based on Kohonen network structure is to improve the quality of the data classification and labeling. New formulation of hexagonal lattice area is used for the enhancement Self Organizing Map structure. The proposed hybrid ESOM/PSO algorithm uses PSO to evolve the weights for ESOM. The weights are trained by ESOM in the first stage. In the second stage, they are optimized by PSO. In the proposed algorithm, the result is measured by using a classification accuracy and quantization error techniques.