Taxi trajectory prediction plays an important role in perceiving urban traffic conditions and analyzing taxi passengers' travel behaviors. In this paper, an urban taxi trajectory prediction model is proposed based on attention mechanism and Long-short Term Memory (LSTM). In this model, road networks is partitioned into grids, and the road segment with the grid which the road segment lies in is represented with embedding vector. Encoder-decoder framework is adopted, and both the encoder and decoder are implemented with LSTM, in which encoder is used to turn the taxi trajectory into coding vector and decoder is employed to transform the coding vector back into taxi trajectory. Furthermore, in order to improve the performance of the trajectory prediction model, attention mechanism is introduced to put the attention on the combination among road segments during prediction. In the experiment, the proposed model was fully verified using Xi'an taxi GPS data. The result shows that the model can effectively predict the city-scale taxi trajectory, and its prediction performance is better than that of the traditional time series prediction models and the existing deep network models.
[1]
Weiwei Sun,et al.
Probabilistic Robust Route Recovery with Spatio-Temporal Dynamics
,
2016,
KDD.
[2]
Yoshua Bengio,et al.
Neural Machine Translation by Jointly Learning to Align and Translate
,
2014,
ICLR.
[3]
John Krumm,et al.
Hidden Markov map matching through noise and sparseness
,
2009,
GIS.
[4]
Xiaohui Yu,et al.
Mining moving patterns for predicting next location
,
2015,
Inf. Syst..
[5]
Jürgen Schmidhuber,et al.
Long Short-Term Memory
,
1997,
Neural Computation.
[6]
Hwasoo Yeo,et al.
Network-Wide Vehicle Trajectory Prediction in Urban Traffic Networks using Deep Learning
,
2018,
Transportation Research Record: Journal of the Transportation Research Board.
[7]
Zhe Zhu,et al.
What's Your Next Move: User Activity Prediction in Location-based Social Networks
,
2013,
SDM.
[8]
Weiwei Sun,et al.
Modeling Trajectories with Recurrent Neural Networks
,
2017,
IJCAI.