神经网络模型用于多变量综合预测

本文研究神经网络用于多变量时间序列预测的原理与方法，提出组合多种信息的综合预测方法。以股票交易为例，用神经网络组合各类信息，运用信心股价理论对中国股市的发展进行跟踪预测。在此基础上进一步从信息利用的角度说明了神经网络预测方法的特点。结果表明，神经网络模型用于多变王时间序列预测，其精度和趋势均较统计方法有所提高；神经网络综合预测，对中短期股票价格的预测，有实用价值。

THE MULTIVARIABLE COMPREHENSIVE FORECASTING WITH NEURAL NETWORKS

In this paper,the principle of neural network forecasting is investigated.Through com-puter simulations,the forecasting ability of neural networks in multivariable time series is compared withthe ARIMA method.The neural network models give out more reasonable results in both accuracy andtrend,In additio.atechnique of comprehensive forecasting with neural networks is proposed,in whichinformation from different sources is gradually combined.Based on these researches,we can predict thestock prices a week ahead of time. In the end,we have concluded that the neural networks make use ofthe information in a more effective way than the common statistical approaches.