基于多任务学习时间卷积网络的路网短时交通速度预测

Short-term traffic speed prediction of road network withmulti-task learning temporal convolutional network

短时路网交通速度预测是智能交通系统的重要技术。基于深度学习,文章提出了一种多任务学习时间卷积网络(Multi-Task Learning Temporal Convolutional Network,MTL-TCN),用于预测路网层面的短时交通速度。以广州市路网交通速度数据为数据源展开对比实验,结果表明MTL-TCN模型优于经典方法和深度学习方法。

Short-term network-wide traffic speed prediction is an important technique to Intelligent Transportation Systems.Based on deep learning,a Multi-Task Learning Temporal Convolutional Network(MTL-TCN)is proposed to predict network-wide short-term traffic speed.With traffic speed data from Guangzhou,the MTL-TCN model is superior to classical methods and deep learning methods.