The Combination Forecasting Model Based On The GM(1,1) Dynamic Equal Dimensional Residual Correction-BP Neural Network

Effective stock forecast is very important in the area of financialinvestment, but stock is always affected by many factors such as policy, economy, psychologies of investors and so on, it is a complicatedtypical nonlinear system which is also hard to forecast accurately. So stock price forecast has become one of the topics in financial market which needs to be studied and resolved urgently.GM(1,1) model and BP neural network model are the useful forecasting models in stock forecast, they are simple, and easy operation.But simplex forecasting accuracy is hard to achieve the required precision from shareholders and investors, so it has little application value.For better raising the forecasting precision that satisfying the effective demands of shareholders and investors, the thesis focus on the following fields:First of all, it introduces the research status and the speculativeknowledge of stock forecast; secondly, it makes research and illustration on the GM(1,1) model, and then does residual revise on the original foundation which raising the precision of model effectively. For better satisfy the actual demands and research needs of theory, the thesisuses the existing stocks to carry on GM(1,1) and BP neural networkcombined model again on the base of residual revise. It uses the residual revise model of GM(1,1) to get the forecast information whichas input data, and also uses the actual daily stock closing prices as target data to get the BP neural network trains, through these trains itbuilds the combined forecasting system. After building the model, the forecasting daily stock closing prices can be received after inputtingthe actual daily data, and then we make error detection on them, beup to the standards, the model works.