一种基于组合神经网络的时间序列预测方法

A TIME SERIES PREDICTION APPROACH BASED ON COMBINED NEURAL NETWORKS

本文探讨了神经网络时间序列预测模型的建立机制及其构造方法。同时，为了消除模型的系统偏差，提出了构造辅助神经网络用以对原有模型的预测结果进行校正以减小其误差。并对外汇汇率数据进行了模型构造和预测。结果表明，组合神经网络在模型的拟合精度和预测准确性方面都有提高。

In this paper, we discuss the constructing method of time series prediction models by neural network. In order to remove the systematic deviations of the models, we propose the method of constructing auxiliary neural network to correct the original predictions. Through constructing models and making predictions for the currency exchange rate data, we can see that the predictions of combined neural networks are improved.