The Method Of Futures Prices Forecasting Research Based On Neural Network

Following the economic development and people's awareness of the investment changes,Futures has become an important component of the modern life.The healthy development and prosperity of the futures markets has also become a concern for investors and managers.In order to effectively play the futures market hedging and price discovery of the basic functions,the risk-control on the futures market study attracted wide attention,because of Price risk all the time was the futures risk control's focus of the study and centers,Therefore,analysis and forecast changes in futures prices will naturally become the trend of the futures market the most important risk-control study.But the structure of system of futures price's complexity and Variability of external factors that determine the difficulty of this task,The traditional forecasting methods can not meet this need.ANN is a kind of intelligent information processing technology,the emergence of neural network to deal with high dimensionality, strong interference difficult to price changes in the complexity of modeling the process of providing an alternative method.In this paper adopted the BP neural network neural network forecasting futures prices,the main research contents of the following four aspects:1.For the traditional BP neural network model of the existence of the operation results by the initial impact,poor stability,slow convergence and easy to fall into local minimum value,such as defects.In this paper,we use three modified BP algorithms to train model,the result shows that improved algorithm for the improvement of the operating speed of more effective,but it can not significantly improve the model stability and the operation of prediction accuracy.2.For the initial weights and threshold parameters'stochastic problem of existing in BP neural network,we advanced genetic algorithm and BP neural network combining to be used the futures price forecasting.The experimental results show that the model fusion algorithm based on the operational stability and prediction accuracy are better than the BP neural network model.However,the possibility of running a "premature" convergence phenomenon may appear.3.For genetic algorithm optimized BP model may appear the "premature" convergence problem,we advanced the particle swarm algorithm and BP neural network to be integrated.The experimental results show that the use of particle swarm optimization algorithm for neural network model to improve the neural network models to predict the stability and prediction accuracy is helpful,but it may also appear a "premature" convergence phenomenon of running.4.Through the MATLAB language proposed in this paper the corresponding futures price forecasting model algorithm,using Chinese copper futures,soybean futures and wheat futures as an example,Verify the improvements made by the neural network model to improve the futures forecast of the feasibility and practicality of algorithms.This study has enriched the futures price forecasting methods,and has important theoretical and practical value for futures prices forecasting.