高级检索
    赖培源, 李程, 王增辉, 王昌栋, 廖德章. 基于图提示微调的交通流量预测[J]. 计算机研究与发展. DOI: 10.7544/issn1000-1239.202440113
    引用本文: 赖培源, 李程, 王增辉, 王昌栋, 廖德章. 基于图提示微调的交通流量预测[J]. 计算机研究与发展. DOI: 10.7544/issn1000-1239.202440113
    Lai Peiyuan, Li Cheng, Wang Zenghui, Wang Changdong, Liao Dezhang. Traffic Flow Prediction Based on Graph Prompt-Finetuning[J]. Journal of Computer Research and Development. DOI: 10.7544/issn1000-1239.202440113
    Citation: Lai Peiyuan, Li Cheng, Wang Zenghui, Wang Changdong, Liao Dezhang. Traffic Flow Prediction Based on Graph Prompt-Finetuning[J]. Journal of Computer Research and Development. DOI: 10.7544/issn1000-1239.202440113

    基于图提示微调的交通流量预测

    Traffic Flow Prediction Based on Graph Prompt-Finetuning

    • 摘要: 交通流量预测是建设智慧城市重要的基础功能,对城市的交通管理和用户出行规划具有重要的意义. 由于时间维度和空间维度的扩展,交通流量的数据具有规模大、快速增长、实时更新等特征,由于时间维度和空间维度的扩展,交通流量的数据具有规模大、快速增长、实时更新等特征,传统的训练模型通常需要将大量的历史数据进行训练预测,导致较长的计算时间和较高的算力成本,因此,如何使用低计算成本的预测模型来满足广泛的流量预测需求,是重要的技术挑战. 近年来兴起的提示微调范式在自然语言处理的下游任务推广中取得了较好的效果,受其启发,提出利用少量的实时数据,来微调优化用大规模历史数据预训练的模型,为交通流量模型预测的优化应用提出了一种新的思路. 通过引入图提示学习来提升传统交通流量预测(traffic flow prediction based on graph prompt-finetuning,TPGPF)模型的泛化能力,在时空多维度下的交通流量图预测模型中,基于历史数据集进行预测模型的预训练,并引入可学习的提示向量,在预训练模型固化的情况下指导预训练的自监督学习模型,以适应新的数据预测任务,提升交通流量预测模型的通用性和有效性. 通过在5个公开数据集上进行了大量的实验,证明了TPGPF的有效性.

       

      Abstract: Traffic flow prediction is a crucial foundational aspect in the development of smart cities, holding significant implications for urban traffic management and user travel planning. The expanding dimensions of both time and space contribute to the increasing volume of traffic flow data, with real-time updates presenting a challenge in deploying cost-effective forecasting models to meet diverse traffic prediction demands. Inspired by the success of the emerging paradigm of graph-based finetuning in downstream tasks of natural language processing, this paper introduces, for the first time to our best knowledge, graph based finetuning to enhance the generalization capabilities of traditional traffic flow prediction models, called traffic flow prediction based on graph prompt-finetuning (TPGPF), which could enhance the generality and effectiveness of self-supervised learning. In the context of spatiotemporal multidimensional traffic flow prediction models, the approach involves pretraining the model based on historical datasets and introducing learnable prompt vectors. With the pretrained model solidified, a self-supervised learning model is guided by the introduced prompt vectors to adapt to new data prediction tasks, thereby enhancing the generality and effectiveness of traffic flow prediction models. Extensive experiments on five real-world public datasets validate the effectiveness of our work. This provides an effective method to overcome the cost challenge brought by quasi real-time training of traffic flow data.

       

    /

    返回文章
    返回