The Application Of BP Neural Network Optimized By Genetic Algorithm In Stock Market Forecasting

Stock, as a product of the market economy, has now become an integral part of the financial markets, playing a very important role in pushing the national economy to a healthy development, satisfying financial capital needs of enterprises and helping social redistribution of wealth and personal financial investment. But stock prices are affected numerous factors, rather like business conditions, the policy trend, economic environment and many others. Investment in the stock market is facing great risk. To ensure the vast majority of small and medium-sized investors a maximized risk and increased revenue, an effective analytical method to assist decision-making is essential and necessary when stock investment activities are under progress.Despite many ways to forecasts the price on the stock market, the traditional forecasting models are mostly based on a long-term, statistical analysis of large amounts of data, which has raised tough requirements of distribution regularity and integrity. Due to the fact that the stock market is a complex multi-variable nonlinear dynamic system, there is a big limitation of the traditional method of forecasting stock price of short-term movements. The artificial neural network has good nonlinear approximation ability as well as the capabilities to handle unclear, disordered and complex information. Its characteristics are proper and appropriate to overcome shortcomings of traditional approaches, thereby reaching a higher accuracy in the short term. Many scholars home and abroad have recently applied artificial neural network prediction into the stock market, and achieved good results and effects.Therefore, this article focuses on widely used, proven mature BP neural network to forecast stock prices. Its relevant mathematical theory, technology and method are discussed in detail; the current situation and the problems caused by its application home and abroad are analyzed. In order to solve the problems that arbitrary initial values of BP network affect the accuracy Problem, we find BP genetic algorithm to optimize BP neural network. After several experiments we find convergence network speed, once optimized, has been improved greatly.To verify the stability and practicality of this algorithm in the experiment, we chose as the experimental samples the Shanghai A shares of Anhui Expressway Company Limited and Sinopec stock, two different types of data. As stock prices are forecast short-term, opening price, ceiling price, bottom price, trading volume and MA5 of three consecutive days are taken as an input sample, while the counterparts of the 4th day as the output sample. Training samples are to be established, then to optimize and finally find out the best individual by choosing, crossing and operating. Comparing the results, we can find Genetic Algorithm optimization BP network predicts more accurate of the two stocks than ever before, also it has a significant improvement of the convergence of training network speed.The findings are that, the Genetic Algorithm has the ability to optimize BP network. The application of BP network model upgraded by genetic algorithm into stock price prediction is feasible and effective. Additionally, it's also found that the algorithm only improves the prediction accuracy of the original BP network, but inefficient to reduce the error of BP neural network. The next stage will combine with other algorithms to achieve better prediction results. And the stability and the maturity of Genetic Algorithm remain to be further improved and verified.