摘要
运用反向传播的改进方法LMBP算法,选取宜昌站1994年～2003年的日流量时间序列,建立LMBP神经网络日流量预测模型.论文对模型中采用不同的延迟时间和不同的激活函数对日流量预测的影响进行了比较和分析.实验结果表明该模型在进行日流量预测时,准确率高,收敛速度快,性能良好.
Levenberg-Marquardt Backpropagation（LMBP） algorithm which is an improvement method of back propagation is applied in daily flow forecast in Yichang station. The effects of different delay times and transfer functions on the forecast are discussed in this paper. The method is evaluated on the data of Yichang station from 1994 to 2003. Experiment results illustrate that the proposed method can pro- vide high accuracy and fast convergence speed for the forecasting of daily flow time series in the station.
出处
《计算机与数字工程》
2013年第11期1733-1735,1790,共4页
Computer & Digital Engineering
基金
海军工程大学青年基金项目(编号:HGDQNJJ13153)资助
