基于BP神经网络改进的黄金价格预测

Gold Price Forecast Based on Improved BP Neural Network

在黄金期货价格预测问题的研究中,价格具有严重的非线性、高噪声和影响因子难以确定等因素,决定了预测的难度。传统方法对黄金价格的预测都强调依赖于黄金价格间的线性关系,局限性明显,导致了预测精度不高。为提高黄金价格预测精度,提出一种投影寻踪优化的BP神经网络改进模型。先通过定性分析得到影响黄金价格波动的因子,然后采用投影寻踪方法选择很强影响力的因子作为神经网络的输入节点,并采用改进的算法进行学习,寻找最优的BP网络结构,利用改进模型,黄金期货价格实现了高精度仿真。结果证明,模型为黄金价格预测提供了一种有效的高精度预测工具。

Gold futures price is the combined result of a large number of factors. Because of high nonlinear,high noise and because the factors is determine difficultly,the prediction is complex and difficult. Traditional methods of forecasting the price of gold have emphasized the intrinsic value of gold,or dependent on the linear relationship between the prices of gold. The limitations are obvious,which leads to the low prediction precision. To further improve the prediction accuracy of the price of gold,an improved BP neural network model based on projection pursuit optimization is proposed. First,we get the factors affecting price volatility of gold using qualitative analysis,and then use projection pursuit to select the strong influence factors as input of neural network nodes,and a modified learning algorithm to find the optimal structure of BP network. Finally,we use computer programming to achieve the gold futures high precision simulation. The experiments show that,the model for the gold price forecast provides an effective tool for accurate prediction.