不同预测算法在简单交通场景中的应用比较

Comparison of Typical Prediction Algorithms in Traffic Forecasting Application

为探讨不同趋势预测算法在简单交通场景中应用的有效性,以部分高速公路收费站数据集为研究对象,分别采用自回归积分滑动平均(Autoregressive Integrated Moving Average,ARIMA)模型、长短期记忆(Long Short-Term Memory,LSTM)循环神经网络和Prophet时间序列预测算法建立交通流预测模型。通过对比分析发现,3种预测模型在解决交通流预测问题方面均表现良好,相比之下,LSTM在模型拟合和预测精度方面表现更好,泛化能力更强,且在影响因素设置方面更为灵活。在以后的研究中,可采用LSTM,结合调参方法解决更多交通场景下的交通流预测问题。

ARIMA(Autoregressive Integral Moving Average)model,LSTM(Long Short-Term Memory)recurrent neural network,and Prophet time series prediction algorithm are used to process the data sets from expressway toll stations to investigate their performances.The results show that the three prediction models all perform reasonably well.But closer study indicates that,comparatively,LSTM performs better in aspects of model fitting and prediction accuracy,and it has stronger generalization ability and is more flexible in influencing factor parameter setting.