基于Conv1D和LSTM组合模型的多步交通流量预测分析

Multi-step Traffic Flow Prediction Based on Conv1D and LSTM Combined Model

为了提高交通流预测精度,设计一种Conv1D和LSTM相结合的多步交通流预测模型。用Conv1D获取交通流的时间与周期参数,并测试各外部因素引起的交通流量变化,再利用LSTM序列分析模型,根据上述交通流特征进行预测。研究结果表明:在加入时间信息的条件下,交通流量预测结果与实际结果间存在更大相似度,考虑时间信息影响能够获得更准确的预测结果。经过外部因素提取处理后可以获得更高精度交通流预测结果,采用Conv1D+LSTM模型可以达到比LSTM更高的精度。

In order to improve the accuracy of traffic flow prediction,a Conv1D and LSTM multi-step traffic flow prediction model was designed.The time and cycle parameters of traffic flow were captured with Conv1D,and the changes of traffic flow caused by external factors were tested.LSTM sequence analysis model was used to predict the above traffic flow characteristics.The results show that when time information is added,there is a greater similarity between the traffic flow prediction results and the actual results.Considering the influence of time information,more accurate traffic flow prediction results can be obtained.After the extraction of external factors,higher precision traffic flow prediction results can be obtained.The Conv1D+LSTM model proposed can achieve higher accuracy than LSTM.