摘要
随着智能出行时代的到来,对出行方式的精准预测是当前学术界和工业界关注的热点问题本文对行人的G P S日志数据进行分割,定义了一些有效且较为容易获取的特征变量,同时,采用基于图的后处理算法来进一步提高出行特征效果,最后采用监督学习中提升树模型进行训练预测,模型预测准度可达76.2%,基于图的后处理对模型精度提升近4%,证明将模型用于对于行人出行方式预测具有较好的拟合度与预测精度,较为准确的预测了用户出行方式的概率分布.
With the advent of the era of intelligent travel,accurate prediction of travel modes is a hot issue of current academic and industrial circles.In this paper,the GPS data of pedestrians are segmented,and some effective and easily acquired feature variables are defined.At the same time,the graph-based-post-processing algorithm is used tofurther improve the travel feature effect.Finally,the Boosting Tree model in supervised learning is used for training.It is predicted that the model prediction accuracy can reach 76.2%,and the graphbased-post-processing improves the model accuracy by nearly 4%.It proves that the model has good fitting and prediction accuracy for pedestrian travel prediction,and the probability distribution of the user's travel mode is predicted more accurately.
作者
李勇
LI Yong(College of Management and Economics,Tianjin University,Tianjin 300072,China)
出处
《综合运输》
2020年第3期22-27,33,共7页
China Transportation Review
关键词
智能出行
监督学习
后处理
精准预测
出行方式
Intelligent travel
Supervised learning
Post-processing
Accurate prediction
Travel mode
