基于ASTLSTM的地铁乘客流量短时预测

Short-term Forecast of Subway Passenger Flow Based on ASTLSTM

地铁乘客流量预测是智能交通系统的重要环节,当前大多数预测模型较少对地铁乘客流量进行时空相关性建模,且未考虑空气质量等天气因素带来的影响,存在地铁乘客流量预测准确度不高的问题。针对以上问题,提出基于注意力机制的时空长短期记忆(ASTLSTM)网络的地铁乘客流量短时预测模型。首先,对数据进行预处理;然后,利用注意力机制与图卷积网络(GCN)、卷积神经网络(CNN)相融合,挖掘地铁数据中的时空相关性,并通过长短期记忆网络(LSTM)来提取空气质量数据中的外部特征;最后,通过特征融合得到地铁乘客流量预测结果。实验结果表明,ASTLSTM模型与LSTM、Conv LSTM等典型模型相比,在短期的地铁乘客流量预测上都有较高的准确度。

The prediction of subway passenger flow was an important part of the intelligent transportation system.Currently,most existing prediction models had limited modeling of the spatiotemporal correlation of subway passenger flow and could not take into account the impact of weather factors such as air quality,resulting in low accuracy in predicting subway passenger flow.To address these issues,a short-term prediction model of subway passenger flow based on the attention mechanism of spatio-temporal long short-term memory network(ASTLSTM)was proposed.Firstly,data preprocessing was performed.Then,attention mechanism was combined with graph convolutional network(GCN)and convolutional neural network(CNN)to mine the spatiotemporal correlation in subway data.External features from air quality data were extracted using long short-term memory(LSTM)network.Finally,feature fusion was performed to obtain the final prediction results for subway passenger flow.The experimental results showed that the ASTLSTM model had higher accuracy in short-term prediction of subway passenger flow compared to typical models such as LSTM and Conv LSTM.