计算机工程与应用2025，Vol.61Issue(11)：67-82,16.DOI:10.3778/j.issn.1002-8331.2407-0410
基于深度学习的短时交通流预测研究综述
Short-Term Traffic Flow Prediction Based on Deep Learning
摘要
Abstract
Traffic flow prediction is an important part of intelligent transportation system,which aims to accurately esti-mate the traffic flow of a specific area in a specific time interval in the future.With the increase of vehicles and the com-plex space-time relationship between different regions in the road network,traditional traffic prediction methods are diffi-cult to accurately describe the characteristics of traffic data,while deep learning prediction methods can better deal with complex feature structures.Therefore,deep learning has become a research hotspot in short-term traffic flow prediction.Firstly,the research status of traditional traffic flow prediction methods and deep learning traffic flow prediction methods is summarized,and the deep learning architecture of convolutional neural network,autoencoder,recurrent neural network,graph convolutional neural network,attention mechanism and Transformer as well as deep learning hybrid neural network is introduced in detail.And the deep learning traffic flow prediction literature,deep learning hyperparameters and scenarios are summarized and analyzed.Secondly,the paper summarizes the common domestic and foreign public data sets in the existing literature.Then,the performance of traffic prediction models is compared and analyzed according to previous model experiments.Finally,the future research direction of traffic prediction based on deep learning is discussed.关键词
交通流预测/深度学习/短时交通流/交通数据集/时空特征Key words
traffic flow prediction/deep learning/short time traffic flow/traffic data set/spatio-temporal characteristics分类
计算机与自动化引用本文复制引用
熊章友,李卫军,朱晓娟,杨国梁,马馨瑜..基于深度学习的短时交通流预测研究综述[J].计算机工程与应用,2025,61(11):67-82,16.基金项目
宁夏高等学校科学研究项目(NYG2024086) （NYG2024086）
中央高校基本科研业务费(2021JCYJ12,2022PT_S04) （2021JCYJ12,2022PT_S04）
国家自然科学基金(62066038,61962001) （62066038,61962001）
北方民族大学研究生创新项目(YCX24363). （YCX24363）
