A New Hybrid Pruning Neural Network Algorithm Based on Sensitivity Analysis for Stock Market Forecast

Considering the design problems of neural network structure, this paper puts forward a kind of hybrid pruning algorithm. The new algorithm mainly contains two steps. First, it combines the Cooperative Coevolutionary Genetic Algorithm (CCGA) and Back Propagation (BP) algorithm to adjust the structure and weight values of neural network; second, it calculates the sensitivity of the hidden-layer neurons which based on the insensitive neurons of the network. Hence, the new hybrid algorithm can not only ensure the generalization capability of the model, but also it can simplify the network structure. Using the improved hybrid pruning algorithm to forecast the stock market and comparing with the GA-BP algorithm, experiment results show this new algorithm has better generalization ability and higher fitting precision to achieve stock forecast.

[1]  Baikunth Nath,et al.  A fusion model of HMM, ANN and GA for stock market forecasting , 2007, Expert Syst. Appl..

[2]  Wichian Premchaiswadi,et al.  Stock Exchange of Thailand Index Prediction Using Back Propagation Neural Networks , 2010, 2010 Second International Conference on Computer and Network Technology.

[3]  Alfred Jean Philippe Lauret,et al.  A node pruning algorithm based on a Fourier amplitude sensitivity test method , 2006, IEEE Transactions on Neural Networks.

[4]  Tommy W. S. Chow,et al.  Self-Organizing and Self-Evolving Neurons: A New Neural Network for Optimization , 2007, IEEE Transactions on Neural Networks.

[5]  Ehud D. Karnin,et al.  A simple procedure for pruning back-propagation trained neural networks , 1990, IEEE Trans. Neural Networks.

[6]  Mitchell A. Potter,et al.  The design and analysis of a computational model of cooperative coevolution , 1997 .

[7]  Feng Li,et al.  Application Study of BP Neural Network on Stock Market Prediction , 2009, 2009 Ninth International Conference on Hybrid Intelligent Systems.