城市多出行方式轨迹预测的模型集成方法

An urban multi-modal travel trajectory prediction method based on model integration

针对轨迹位置预测研究中出现的数据稀疏性和来源单一性问题,提出了一种基于长短期记忆网络的模型集成算法。该算法结合公交、滴滴、共享单车等多出行方式的轨迹数据,针对每种出行方式的特点,利用LSTM模型,计算最佳预测模型参数,进行加权平均集成,预测下一时刻区域交通的轨迹位置。以北京市朝阳区为研究区域,进行实验分析,实验结果表明:该文所提供的LSTM模型集成算法不仅解决了数据稀疏性和单一性问题,提高了模型预测的精度,还能更好地预测轨迹对象下一时刻的位置,反映城市中车辆的行驶趋势。研究结果可以对位置服务、城市交通管制以及规划提供一定的建议。

Aiming at the problems of data sparsity and single source arising in trajectory location prediction research,the paper proposes a model integration algorithm based on long and short-term memory network.The algorithm combines the trajectory data of multiple travel modes,such as public transportation,DiDi,and bike-sharing,and for the characteristics of each travel mode,it uses the LSTM model,calculates the optimal prediction model parameters,and carries out the weighted average integration to predict the trajectory position of regional traffic at the next moment.This study takes Chaoyang District of Beijing as the study area and conducts experimental analysis,and the experimental results show that the LSTM model integration algorithm provided in this paper not only solves the problems of data dilution and singularity and improves the accuracy of model prediction,but also better predicts the position of the trajectory object in the next moment,reflecting the trend of the vehicles traveling in the city.The results of the study can provide certain suggestions for location services,urban traffic control,and planning.