Stock Market Prediction Research Based On Artificial Neural Network

Bond investment has become increasingly important part of social life along with the development of the economic system reform and financial system reform in China. As a kind of bond investment, stock trade is a common risk investment activity in modern economic life. Stock market is a complex non-linear dynamic system, which is very difficult to develop the inherent rules using the traditional time series prediction models. In recent decades, there has been a growing interest in applying artificial neural network for predicting further stock behavior.In this paper, it analyses the predictability of stock market and the main four methods for stock price prediction, including investment analysis, time series, non-linear system analysis and combination forecasting method. Then a GA-Elman stock price prediction model is founded based on the Elman artificial neural network. Furthermore, the genetic algorithm is applied to make weight evolution computation in the training algorithm in order to avoid plunging the local minimum. Practical test demonstrates that the GA-Elman stock price prediction model gets higher precision, steadier prediction effect and more rapid convergence speed.To check its feasibility and validity, this paper is divided into three main stages. In the first stage, it is tested with data from China Unicom and Sichuan Changhong. Practical test results show it achieves an overall high coincidence ratio, which presents it predicts the stock price trend accurately in most cases. In the second part, it is compared with BP and RBF stock price prediction models respectively. Experimental results demonstrate that the GA-Elman stock price prediction model outperforms the other two prediction models obviously. In the last stage, it is compared with the time series prediction model in the terms of error evaluation. Analysis results indicate that they have both merits and demerits respectively. However, the GA-Elman stock price prediction model outperforms in aspect of coincidence ratio.Theoretical analysis and practical result show that GA-Elman stock price prediction model is feasible and valid and has a favorable applicable foreground.