基于改进BP神经网络的公交线路OD矩阵推算方法

公交线路OD矩阵是公交线路优化设计调整的基础性数据，针对传统解析模型对输入数据敏感性高造成OD推算效果不理想的缺陷，提出了以改进BP神经网络为基础，根据公交站点上下客数据建立公交线路OD矩阵的推算模型，并设计了OD推算神经网络。进而基于算例数据，应用数值试验对神经网络设计的核心参数进行了标定并给出OD推算的计算结果。算例分析结果表明，使用此方法能显著提高预测的精度，推算结果更具可靠性且更能反映公交线路的需求分布状况。

Estimation a Transit Route O-D Matrix Using On/Off Data: An Application of Modified BP Artificial Neural Network

A route O-D matrix is the base data for transit line designing and optimizing.This paper presents an OD estimation model,which applies a modified BP artificial neural network(ANN) to track the weights of OD pairs within a transit route using the number of boarding and alighting passengers at each stop.This paper details the key issues associated the(estimation) process,including constraints for OD estimation models,design of ANN,the learning and stopping rules,and the calibration of model parameters through computational experiment.The results of a rigorous validation with on/off data from a real bus route reveal that the proposed model is quite effective and reliable in estimating the OD matrix for identification of the underlying demand pattern of a transit route.