Abstract: The essence of fare clearing is the distribution problem of passenger flow in different lines. Thus, a kind of passenger clearing model for rail transit based on BP neural network was proposed, with the full consideration of passenger travel route choice on the basis of multi-factor.The multiple factors that influenced route choice of passengers were route choice were divided into quantified and uncertain factors. By training the inhibitory coefficient and the excitation coefficient of the sample, the results were transmitted to the output layer to output by combining with the transformation function.Compared with the traditional model, the proposed model was more consistent with the multi-factor psychology of passenger travel choice. Finally, by comparing the actual investigation of passenger flow and Logit model, it showed that without the interference of other factors, the proposed method can realize the passenger clearing in different lines well.

Key words: traffic engineering, urban traffic, passenger clearing, neural networks, railway transportation

摘要： 轨道交通费率清分的实质是在不同线路下网络客流分布的问题。在充分考虑乘客出行路径选择多要素的基础上，提出一种基于神经网络的轨道交通 客流清分模型。将影响乘客出行路径选择的多要素分为确定性要素和不确定性要素，通过样本训练神经元的抑制系数和激励系数，结合转换函数将结果传 导给输出层输出。与传统模型相比，该模型更符合乘客出行选择的多要素心理。最后通过对比客流调查结果和Logit模型表明，在排除其他要素的干扰下 ，该方法能够较好的实现客流在不同线路的清分。

关键词: 交通工程, 城市交通, 客流清分, 神经网络, 轨道交通