基于神经网络的期货预测数据预处理问题研究

期货预测研究在期货价格数据预处理和预测方法上存在直接套用原始数据代入模型以及价格预测模型和原始数据模型不相匹配等问题,需要予以解决.本研究在采用通货膨胀率指数调整、平均周期项以及滤波等方法对铜期货价格时间序列数据进行预处理后,分别将预处理前后的期货价格数据输入到神经网络预测模型,通过比较两者预测结果来验证原始期货时间序列数据预处理的必要性.

Studies of Data Processing in Future Forecasting Using Neural Networks Method

At present,there are some mistakes in pretreatment and forecasting of time series data of future price in some researches,such as using raw data indiscriminately and improper model corresponding.This paper used inflation rate adjustment,moving-average method and stochastic filter to pretreatment raw copper future price data series,then input the raw data and processed future price time series data into neural networks forecasting model.Comparing the forecasting results between raw data and processed data,the latter effect is better and the average deviation remarkably reduced.