The invention relates to an urban traffic data prediction method based on space-time causal prediction of continuous learning, belongs to the technical field of urban space-time data prediction, and solves the problems of inaccurate urban traffic data prediction, too many time and hardware resources required for training and prediction, low prediction efficiency and the like in the prior art. And the actual application requirements of urban traffic data prediction are difficult to meet. Task data is real-time data of multiple batches of tasks, a space-time historical memory library is constructed, time sequence data and graph data which can represent old task data distribution are screened out in a classified mode, the forward migration requirement is met, and the prediction effect is accurate; according to the method, an orthogonal projection operator is calculated from a space-time historical memory bank by combining an orthogonal parameter space of a model weight, parameters are updated in the orthogonal parameter space during training on a new task, the prediction effect on an old task is stable, past urban traffic data are efficiently utilized, time and hardware resources required by training and prediction are saved, and the prediction efficiency is improved. And the actual application requirement of prediction is met.
本发明涉及一种基于持续学习的时空因果预测的城市交通数据预测方法,属于城市时空数据预测技术领域,解决了现有技术对城市交通数据预测不准确、训练及预测所需的时间和硬件资源过多,难以满足对城市交通数据预测的实际应用需求的问题。本发明的任务数据为多批次任务的实时数据,构建时空历史记忆库,分类筛选出能代表旧的任务数据分布的时序数据和图数据,满足前向迁移需求,预测效果准确;结合模型权重的正交参数空间,从时空历史记忆库中计算出正交投影算子,实现在新任务上训练时在正交参数空间中更新参数,在旧任务上的预测效果稳定,高效利用过去的城市交通数据,节省训练及预测所需的时间和硬件资源,满足预测的实际应用需求。
Urban traffic data prediction method based on space-time causal prediction of continuous learning
基于持续学习的时空因果预测的城市交通数据预测方法
2023-09-12
Patent
Electronic Resource
Chinese
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
European Patent Office | 2022
|Traffic flow prediction method based on space-time causal attention network
European Patent Office | 2024
|European Patent Office | 2024
|Urban road traffic flow prediction method and device based on space-time deep learning model
European Patent Office | 2020
|Urban area traffic flow prediction method considering multiple space-time granularities
European Patent Office | 2024
|