【摘要】 居民交通出行活动研究是制定城市交通政策和规划的重要基础。考虑了出行活动一系列的选择过程,将出行活动和交通方式按照发生序列创建组合链,组合链包含了出行目的、交通方式、活动序列等信息。通过组合链可以得到购物出行频率和交通方式指标。以交通出行需求条件弹性较大的购物出行为基础,利用居民一日出行活动日记数据,提取购物出行组合链。利用神经网络并行处理及非线性拟合的优势构建组合出行链预测模型,分别建立了基于传统神经网络、LM神经网络和切比雪夫神经网络的仿真模型,输入出行者社会经济属性、家庭特征等17个影响因素对组合链模型进行训练,经验证切比雪夫模型的预测效果最优。标定后的模型可以较好的预测购物出行频率和交通方式选择,可为交通需求管理政策研究、效果预判等提供量化分析工具。更多还原

【Abstract】 Research of travel behavior is important for urban transport policy making and planning. From the travel selection process, an activities and mode share integrated-trip-chain of shopping trip was established. The integrated-trip-chain included the information about the occurrence, sequence and mode share of shopping travel activities. Shopping trip frequency and mode share could be calculated by the integrated chain. Three neural network models were selected as the tools to build the models for its parallel processing advantages of simulation and nonlinear fitting. 17 factors including traveler’s social and economic properties, family feature, travel characteristics and etc. were considered as the input of the model, the output was the integrated-trip-chain. From the model training and test results, it can be inferred that the Chebyshev neural network works best. The calibrated Chebyshev neural network model can predict shopping trip frequency and mode share, it can provide a quantitative analysis tool for traffic demand management policy research, the pre-judgment and so on.更多还原