基于径向基神经网络的铁路客货运量预测研究

发布时间：2018-06-06 22:56

本文选题：铁路运输 + RBF神经网络　；参考：《铁道科学与工程学报》2014年04期

【摘要】：根据径向基神经网络具有分析非线性动态系统的混沌特性的特点,对铁路客货运发送量相关时间序列进行分析和研究,在Takens相空间重构的基础上,利用互信息方法求嵌入时延、伪邻域方法求嵌入维数;应用G-P方法和最大Lyapunov指数方法对铁路客货运量时间序列进行混沌识别;根据RBF神经网络的学习算法和辨识原理,对铁路客货运量预测流程进行分析。应用径向基神经网络对铁路客货运量自1999-01-01~2012-08-27共4 988 d的发送量为基础进行径向基神经网络预测;并对预测误差进行检验及对预测结果进行分析。研究结果表明:基于径向基神经网络预测值能很好地与实际值相吻合,因而在铁路客货运量相关时间序列中预测有广泛的实用价值。
[Abstract]:According to the characteristic of the radial basis function neural network (RBF) which has the characteristic of analyzing the chaos characteristic of the nonlinear dynamic system, the correlation time series of railway passenger and freight transport quantity is analyzed and studied. Based on the reconstruction of Takens phase space, the embedded time delay is solved by mutual information method. The pseudo-neighborhood method is used to calculate the embedding dimension, the G-P method and the maximum Lyapunov exponent method are used to identify the time series of railway passenger and cargo traffic, and the prediction flow of railway passenger and freight volume is analyzed according to the learning algorithm and identification principle of RBF neural network. The radial basis function neural network (RBF) was applied to forecast the passenger and cargo transport volume from 1999-01-01g to 2012-08-27. The radial basis function neural network was used to forecast the railway passenger and cargo transport volume from 1999-01-01g to 2012-08-27, and the prediction error was tested and the prediction results were analyzed. The results show that the predicted value based on radial basis function neural network is in good agreement with the actual value, so it has a wide range of practical value in the correlation time series of railway passenger and cargo traffic.
【作者单位】：中国铁道科学研究院电子计算技术研究所;铁道部信息技术中心;中国铁路总公司计划统计部;
【基金】：铁道部科技研究计划资助项目(2007X008-G,2008X015-H)
【分类号】：U293.13;U294.13;TP183

【相似文献】