An optimized wavelet neural network model for stock market prediction integrated with genetic algorithm

Abstract

Stock market is a highly volatile domain. Actually, it has always been a challenge to researchers over the world. In recent years, intelligent methodology like wavelet neural network has been employed in this field. However, wavelet neural network based on the theory of the backpropagation (BP) algorithm has two prominent vulnerabilities: low convergence rate and the easily converging to local minimum point. Thus this project establishes an optimized stock price prediction model based on genetic algorithm and wavelet neural network. This model integrates time-frequency localization of the wavelet neural networks and the global optimization searching performance of genetic algorithm. The results prove that this model can substantially improve the forecasting precision performance than other traditional artificial neural network and wavelet neural network. Furthermore it can also avoid the intrinsic defects of the BP algorithm.

To validate prediction performance, the result has been compared with 3 research papers, “Stocks Market Modeling and Forecasting Based on HGA and wavelet neural networks” written by Zhou and Wei [11] as well as “Wavelet Transform, Neural Networks and The Prediction of S&P Price Index: A Comparative Study of backpropagation Numerical Algorithms” [32] and “A Comparative Study of backpropagation Algorithms in Financial Prediction”[31] written By Lahmiri. Results show that, our optimized system demonstrates outstanding performance over other systems and achieve very good prediction accuracy in terms of both value and direction.