基于个体记忆效应和距离效应的出行目的地识别
Trip destination recognition based on individual memory effect and distance effect
郑思静 1陈勇 2朱奕璋 1陈喜群2
作者信息
- 1. 浙江大学 工程师学院 智能交通研究所,浙江 杭州 310058
- 2. 浙江大学 建筑工程学院 智能交通研究所,浙江 杭州 310058
- 折叠
摘要
通过分析个体出行轨迹数据,挖掘个体出行历史记忆特征以及个体所处位置与潜在目的地的距离特征,建立新的出行目的地识别模型.利用杭州市 200 个匿名个体 62 880 次出行数据测试所建模型的效果.对基于位置的服务(LBS)数据进行预处理,提取以活动为目的的分段出行数据片段,采用GeoHash网格编码方法得到网格化后的个体历史目的地集合.利用随机缺失的个体出行历史轨迹数据构建训练集和测试集,采用非线性最小二乘法对模型进行参数标定.结果表明,所建模型提升了出行目的地识别精度;对比不同模型的召回率、折扣累计收益和F1 分数,所建模型优于马尔可夫模型、决策树模型以及随机森林模型;数据缺失率敏感性分析结果验证了所建模型的鲁棒性.
Abstract
A new travel destination recognition model was established through analyzing individual travel track data,mining the characteristics of individual travel history memory and the distance between individual location and potential destination.The model was tested using 62 880 trips of 200 anonymous individuals in Hangzhou.The location based service(LBS)data was preprocessed,the segmented travel data fragments were extracted for the purpose of activity,and the meshed individual historical destination set was obtained by GeoHash grid coding method.The training set and test set were constructed by using the random missing individual travel history track data,and the parameters of the model were calibrated by nonlinear least square method.Results show that the proposed model improves the recognition accuracy of travel destination.Comparing the recall rate,discount cumulative return and F1 score of different models,the proposed model was better than the Markov model,decision tree model and random forest model.The robustness of the proposed model was verified by the sensitivity analysis of data missing rate.
关键词
智慧出行/个体出行/目的地识别/记忆效应/距离效应Key words
smart mobility/individual travel/destination recognition/memory effect/distance effect引用本文复制引用
基金项目
国家自然科学基金(72171210)
浙江省自然科学基金(LZ23E080002)
出版年
2024
NETL
NSTL
万方数据