Prediction System Of Stock Index Based On Bp Neural Network

With the development of economic, stock market has already become an integral part of the financial markets since it was built. However, stock market is affected by many factors, such as national policy, economic environment, emergencies and manipulation, which make investors to take a huge risk at the same time enjoy the big profit. For the majority of small and middle investors, if the trend of stock market could be predicted, then they can maximize the profit and minimize the risk. In stock market, stock index reflects the overall trend of stock market; hence it has important theoretical and practical significance to predict stock.With analysis of stock market prediction, we learned that the stock market is a complex nonlinear dynamic system, the prediction result of traditional linear prediction method is not ideal, while BP neural networks has strong nonlinear approximation ability, self-learning ability and adaptive capacity, hence BP neural network has become one of the most widely used methods in the field of stock market prediction.This article makes functional requirements and nonfunctional requirements analysis of BP neural network based stock index prediction system, on the basis of requirement analysis, make architecture design and functional design of the system according to design objectives and principles, and carry out a detailed functional and database design. As the core module of this system is stock index prediction module, this article has made experimental analysis of model design and parameter selection of BP NN and enhance BP NN with optimization of the network topology, adding momentum and improvement of the activation function to address some problems in order to improve system performance and forecasting accuracy.Finally, the prediction results show that the stock index prediction system achieve the advantages of fast convergence and high prediction accuracy, with a high practical value. However, the prediction system still exist some shortcomings, such as the prediction scope is too narrow, only can predict the Shanghai Composite Index; prediction accuracy still can be improved. The next phase of work of this article is to study how expand the scope of perdition of this system, and how to combine other algorithms to improve BP neural network prediction model in order to obtain better prediction results.