基于变窗口神经网络集成的时间序列预测

Time Series Forecasting Based on Variable-window Neural Networks Ensemble

提出一个变窗口神经网络集成预测模型。该模型利用自相关分析构造出差异度较大的个体神经网络,提高了预测系统的泛化能力,同时能够有效剔除异常序列,提高预测精度。采用真实世界的数据集对该模型进行仿真。实验结果表明,该预测模型具有较高的预测精度,能有效预测移动通信的话务量。

A variable-window neural network ensemble model is proposed, which takes use of the self-correlation analysis method to construct all the individual neural networks with different types. This model improves the generalization ability of forecasting system. It can also remove outlier series effectively and promote the accuracy of forecasting. The model is simulated by using real data sets. Experimental results show this forecasting model has higher accuracy of forecasting and can predict the traffic of mobile communication effectively.