Elliott waves, Backpropagation neural network, Levenberg–Marquardt adaptation, Pseudo neural network, Analytic programming, Differential evolution

This article presents a comparative study of the classification of Elliott waves in data. Regarding the methods of classification, the paper deals with three approaches. The first one is a multilayer artificial neural network (ANN) with sigmoid activation function and backpropagation and Levenberg–Marquardt training algorithm. Second approach uses training algorithms of ANN but forms of activation functions of hidden nodes and nodes in output layers have been proposed by analytical programming with the differential evolution. The last approach offers results performed by synthesized pseudo neural networks where the symbolic regression is used for synthesis of a whole structure of the classifier, i.e., the relation between inputs and output(s) similar to ANN. In this case, meta-evolution version of analytic programming with differential evolution is used. In conclusion, all results of this experimental study were evaluated and compared mutually.