Gold Futures Forecasting Research Based On Elman Neural Network

Since the gold futures was official listed by SHFE in the year2008, goldfutures has became more and more important in investment and hedging as a tradevarieties.But the fluctuations of prices on gold futures is a very complex system, sousing a model which is based on econometrics for researching gold futures marketis difficult. This paper proposes a method that use neural network for forecastingthe gold futures. But the traditional BP neural network has many defects and lackson forecasting gold futures. This paper has analysed these defects and lacks, andsupposed a solution: using Elman neural network to structure the model of goldfutures forecasting.This paper has structured the model of gold futures forecasting by usingElman neural network, and used the real datum to check the effect of the model,and then, gets the conclusion that the gold futures forecasting model based onElman neural network can forecast the gold futures better. The major contributionshave listed as follows:(1) Supposed10elements, which affect the fluctuations in prices of goldfutures, by analysing the attributes of futures and gold futures.(2) Has introduced the elementary knowledge of neural network andElman neural network, and analysed the defects and lacks of traditional neuralnetwork, and then supposed the model of forecasting based on Elman neuralnetwork.(3) Because when we are structuring a model based on neural network, wewill find that the more dimensionalities, we will need the more computer capacity.So we will use GPPM to confirm the weight of every elements, and reduce theamount of the nodes at the input layer of the model.(4) At last, we structure the model successful, and forecasted ten days’closing price of gold futures in SHFE. And proved the effect of this model which isstructured based on Elman neural network.