融合手机信令数据的轨道交通出行分担率模型研究

交通行为模型广泛应用于城市出行需求分析等领域。传统行为模型的参数设置通常依赖经验判断，模型预测精度缺乏大样本验证手段。本文以重庆市解放碑-观音桥组团通道出行行为为研究对象，融合手机信令数据、AFC数据和问卷调查数据，构建随机参数分别为正态分布、均匀分布和 γ 分布的混合Logit模型，将手机信令数据与AFC数据分析结果作为分担率标杆数据进行模型精度对比，其识别的组团间全天轨道出行分担率为37.13%，当混合Logit随机参数为正态分布时，模型预测的分担率为39.5%，预测精度最高。研究表明，利用手机信令数据等多源数据分析校验传统行为模型精度，定量分析并优选最佳的参数分布形式具有实际意义，能够对提高传统行为模型的预测精度提供借鉴。

Rail Transit Trip Sharing Rate Model Combining Mobile Phone Signaling Data

The transportation behavior model is widely used in urban travel demand analysis and other fields. The traditional behavior model's parameter setting is usually dependent on empirical judgment, with large sample data verification lacking. This paper's research object is the travelers' corridor travel choice behavior between the Jie Fangbei and Guanyin Bridge group in Chongqing. The mixed Logit model with different random parameters is constructed based on the mobile phone signaling data, AFC data, and questionnaire survey data. Meanwhile, the random parameters include normal distribution, uniform distribution, and γ distribution. Then the analysis of cellular data and AFC data are used as benchmark data to optimize the mixed Logit model. The one-day inter-group railway travel sharing rate identified by big data is 37.13%. And the mixed Logit model accuracy reaches the highest (39.5%) when the mixed Logit random parameter is a normal distribution. The result shows that it is practical to use extensive data analysis to verify traditional behavior models' accuracy, quantify and optimize the best parameter distribution. This paper provides ideas for improving the prediction accuracy of the traditional behavior model.