Study On The Stock Exchange Prices Forecast Model

The stock market is a natural outcome of the Market Economy, and it is of great importance to our country's financial field. With its gradual standardizing, people's investment consciousness has enhanced. Nowadays, in the capital market security is no less important than banking and insurance. Stock investment has become an important means of family or individual financing. However, as is known to all, the stock market is characterized by high risk and high income. Thus, the rise and fall of stock exchange prices and its changing trends are the major concern of the government and the common investors.The thesis is mainly devoted to the following aspects:(1)It has conducted a deep study of the factors which affect the stock exchanges prices,the foundation and approaches of stock exchanges prices forecast,and the theory of stock price forecast based on BP neural network.(2)By using forward single-hidden-layer neural network to create stock forecast model, it discusses the problems of how to ascertain numlayers,dimensions of layers and initial parameters.In order to raise the network performance,the author divides data collection into training and testing collection, and normalize inputs,and adopts the improved BP algorithm to train network.(3)Take Minsheng Bank as an example,It has adopted the mobile time window and the technical indicators separately to train created forecasting model, and then use the trained network to forecast the stock data, which works well.According to the inputweights of trained network, it compares impact of each input variable.(4)To validate the applicability of the Stock Exchange Prices Forecast Model, the author selects another five stocks to validate. It turn out to be very effective,which proves the model applicable to other stock.The stock market is a very complicated nonlinear dynamic systems.Yet the nerual network is characterized by approachability to non-linear function, self-learning, and adaptive.As is proved, by analyzing the Past historical data, the BP nerual network is able to find out the inner principle of the stock development, and save it in weights and biases for the use of estimating the future trends.