基于行程数据的公交车到站时间预测

为向乘客提供较为准确的上下车时间参考，解决长距离预测中误差累积明显的问题，构建基于双层、双注意力、双向长短期记忆(LSTM)神经网络的公交车到站时间预测模型，提出一种基于行程数据的公交车到站时间预测方法.以广州市B2路、 560路公交车工作日的实际运行数据为例，对该预测方法进行精度验证.结果表明，由该模型所预测的行程时间，其平均绝对百分比误差为8.09%,在长距离到站时间估算上，15个站点的预测误差可保持在4.00 min左右.

Bus arrival time prediction based on trip data

In order to provide passengers with a more accurate reference for getting on and off the bus and solve the problem of obvious error accumulation in long-distance prediction, this paper constructs a bus arrival time prediction model based on dual attention and bidirectional double-layer LSTM(AALSTM), and proposes a bus arrival time prediction method based on trip data. First, through the attention mechanism, the differences in the impact of each influencing factor on the bus travel time at different times are distinguished; secondly, the self-attention and LSTM are integrated to predict the bus travel time; finally, based on the estimated travel time and The driving rules between bus stops are obtained, and the estimated arrival time of the bus is obtained. Taking the actual operation data of B2 and 560 buses in Guangzhou as an example, the accuracy of the prediction method is verified. The results show that the average absolute percentage error of the travel time predicted by the AALSTM model is 8.09%, and the relative accuracy of the AALSTM, LSTM and MLP models without attention is improved by 0.51, 0.64, and 1.46 percentage points, respectively. Therefore, the prediction method has higher prediction accuracy and better applicability.