Evolved neural network based intelligent trading system for stock market

Abstract

In the present study, evolved neural network is applied to construct a new intelligent stock trading system. First, heterogeneous double populations based hybrid genetic algorithm is adopted to optimize the connection weights of feedforward neural networks. Second, a new intelligent stock trading system is proposed to generates buy and sell signals automatically through predicting a new technical indicator called medium term trend. Compared to traditional NN, the new model provides an enhanced generalization capability that both the average return and variance of performance are significantly improved. © 2013 Springer-Verlag Berlin Heidelberg.