基于手机信令HMM轨迹映射的交通方式识别算法

Transportation mode recognition algorithm based on mobile phone signaling HMM trajectory mapping

为解决手机信令数据稀疏性限制和重构轨迹特征提取与融合能力不足问题,提出一种出行轨迹重构与多源特征融合的交通方式精准识别算法。构建刻画基站信号传播路径损耗与信号强度的无线信号损耗模型,利用隐马尔可夫模型(hidden markov model,HMM)将手机信令轨迹由基站序列重构为路段节点序列,提出基于无线信号传播隐马尔可夫模型(wireless signal propagation hidden markov model,WP-HMM)的出行轨迹重构方法,用以描述信号强度与距离作用关系。基于出行重构后的轨迹,结合路段类型特征,提出了时空标准化相似性度量算法,以融合导航轨迹特征,并构建了基于随机森林(random forest,RF)的交通方式识别算法。实证分析表明:通过出行轨迹的重构,模型的平均识别精度提高了8%以上,且对新样本具有优异的泛化能力;相较于现有方法,时空标准化相似性度量算法能更准确捕捉轨迹间的移动模式;在不同环境下的轨迹识别中,模型在郊区区域的表现显著高于城区。所提算法在大规模手机信令数据的出行方式识别领域具有重要的应用价值。

To address the limitations of sparse mobile phone signaling data and the insufficient ability to extract and fuse features in trajectory reconstruction,this paper proposes a precise transportation mode recognition algorithm based on trajectory reconstruction and multi-source feature fusion.A wireless signal loss model is developed to characterize the signal path loss and signal strength of base stations.Using the hidden markov model(HMM),the mobile phone signaling trajectory is reconstructed from a base station sequence to a road segment node sequence.A trajectory reconstruction method based on the wireless signal propagation hidden markov model(WP-HMM)is proposed to describe the relationship between signal strength and distance.Based on the reconstructed travel trajectory and combined with the characteristics of road segment types,a spatiotemporal standardized similarity measurement fusion navigation trajectory feature is proposed,and a transportation mode recognition algorithm based on random forest(RF)is constructed.Empirical analysis shows that,through the reconstruction of travel trajectories,the model′s average recognition accuracy improved by over 8%,and it demonstrated excellent generalization ability for new samples.Compared to existing methods,the spatiotemporal normalized similarity measure more accurately captures the movement patterns between trajectories.In trajectory recognition across different environments,the model performs significantly better in suburban areas than in urban areas.The proposed algorithm demonstrates significant application value in the field of transportation mode recognition using large-scale mobile phone signaling data.