融合多因素的短时交通流预测研究

交通流预测一直是交通领域的研究热点，针对现有交通流预测研究大多为常态下的预测，而未考虑天气、节假日等外部因素的影响，提出了一种融合多因素的短时交通流预测模型。通过长短时记忆网络（long short-term memory，LSTM）捕捉时间序列的长期依赖关系，引入注意力机制，利用注意力机制自适应地选择相应的驱动序列，实现短时交通流的预测。实验分别与传统模型、未引入注意力机制的CLA-ATTN模型及未融合多因素的CLA-MFACTOR模型进行对比分析，结果证明所提出的CLA模型具有较高的预测准确度，是一种较好的预测方法。

Short Term Traffic Flow Prediction Based on Multi-Factors

Traffic flow prediction has always been a hot research topic in the field of transportation. In view of the fact that most of the existing traffic flow prediction researches are under normal conditions without considering the influence of weather, holidays and other external factors, this paper proposes a short-term traffic flow prediction model integrating multiple factors. Long short-term memory（LSTM） is used to capture the long-term dependence of time series, and then the attention mechanism is introduced to adaptively select the corresponding driving sequence to realize short-term traffic flow prediction. By comparing with the traditional model, the CLA-ATTN model without attention mechanism and the CLA-MFACTOR model without the introduction of multi-factors, the results show that the CLA model has higher prediction accuracy and is a better prediction method.