论文部分内容阅读
针对传统BP神经网络模型存在的计算效率和泛化能力低的问题,采用双层次特征分析方法对铁路旅客发送量统计数据的时间特征进行分析,提取出日趋势特征、月趋势特征、日周期性特征、月周期性特征、春运-暑运特征和黄金周-小长假特征作为模型的输入变量,建立双层次的BP神经网络模型,然后根据Gram-Schmidt正交化定理对双层次BP神经网络模型进行改进,在隐含层的输出采用Gram-Schmidt变换增加投影层,从而得到双层次正交神经网络模型。该模型包括2个相对独立的网络模型,1个用于处理客运量日数据,另1个用于处理月数据,2个网络模型的输出经过合成,最终得到客运量的预测结果。模型的应用证明,在铁路客运量预测中双层次正交神经网络模型比传统的BP神经网络模型更为有效。
Aiming at the problem of low computational efficiency and generalization ability of traditional BP neural network model, the time characteristic of railway passenger throughput statistics is analyzed by bilevel eigenanalysis method, and the daily trend characteristics, monthly trend characteristics, daily cycle Sexual characteristics, monthly periodic characteristics, Spring Festival - Shuyun characteristics and Golden Week - small holiday features as input variables of the model to establish a two-level BP neural network model, and then according to the Gram-Schmidt orthogonalization of bilevel BP The neural network model is improved, and the output layer of the hidden layer is enhanced with the Gram-Schmidt transform to increase the projection layer. Thus, a bi-level orthogonal neural network model is obtained. The model includes two relatively independent network models, one is used to process the daily data of passenger traffic, the other one is used to process the monthly data, and the outputs of the two network models are synthesized to finally get the forecast of passenger traffic. The application of the model proves that the bi-level orthogonal neural network model is more effective than the traditional BP neural network model in the railway passenger volume forecasting.