The invention discloses a time sequence short-term and temporary prediction method, and relates to the technical field of spatio-temporal data prediction.The method comprises the steps that road traffic flow and time information are extracted from GNSS data of a vehicle to serve as original input, then a DTW algorithm is introduced to conduct fine adjustment on time features of the original input, so that new time features can be better matched with historical data, and the accuracy of time sequence short-term and temporary prediction is improved. And then the LSTM network with an attention mechanism is used to extract the feature vectors for prediction, the introduction of the attention mechanism can balance the influence degrees of different time periods, and finally the traffic speed is predicted through a traffic speed prediction model. According to the invention, the DTW algorithm and the attention mechanism are introduced to improve the prediction capability of the LSTM network under different features, and the data after fine tuning is obviously improved for holidays and festivals; according to the method, the DTW algorithm is introduced to carry out fine adjustment on the time characteristics, so that the current data distribution is closer to the historical data distribution, and holidays and festivals can be better predicted without adding extra information due to the introduction of the DTW algorithm.
本发明公开了一种时间序列短临预测的方法,涉及时空数据预测技术领域,从车辆的GNSS数据中提取道路交通流和时间信息作为原始输入,然后引入DTW算法对原始输入进行时间特征上的微调,使得新的时间特征能够更好的和历史数据相匹配,再利用具有attention机制的LSTM网络提取特征向量进行预测,attention机制的引入可以权衡不同时间段的影响程度,最后通过交通速度预测模型预测交通速度。本发明引入DTW算法和attention机制提高LSTM网络在不同特征下的预测能力,对于节假日使得微调后的数据得到明显的改善;引入DTW算法对时间特征进行微调,从而使得当前数据分布与历史数据分布更加接近,且由于DTW算法的引入,可以在不增加额外信息的情况下对节假日进行更好的预测。
Time sequence short-term and temporary prediction method
一种时间序列短临预测的方法
2023-07-04
Patent
Elektronische Ressource
Chinesisch
Road visibility short-term and temporary prediction method based on ensemble learning
Europäisches Patentamt | 2020
|Short-term travel time prediction
Elsevier | 2003
|Short-term travel time prediction
Online Contents | 2003
|Tema Archiv | 2007
|Reliable Real-Time Framework for Short-Term Freeway Travel Time Prediction
Online Contents | 2006
|