The invention discloses a federated transfer learning algorithm based on privacy protection in cross-city traffic flow prediction, and provides a multi-modal graph convolutional network and time domain convolutional network model 2MGTCN. And the accuracy of the TFP is ensured while the data privacy is protected. The model comprises three main components: a source domain, transfer learning and a target domain, and in the model, a gray relation analysis method and a dynamic time warping method are adopted to capture a road relation and a flow relation respectively. In addition, spatial and temporal information of the road network is captured using a graph convolutional network and time domain convolutional network architecture. In order to protect privacy of target city data, federal learning is used to train a target model, and a parameter aggregation algorithm based on spatial similarity is provided.
本发明公开了跨城市交通流预测中一种基于隐私保护的联邦迁移学习算法,提出了一个多模态的图卷积网络和时域卷积网络模型2MGTCN。在保护数据隐私的同时,保证TFP准确性。该模型包含三个主要组成部分:源域、迁移学习和目标域,在该模型中我们采用灰色关系分析方法和动态时间规整方法分别捕获道路关系和流量关系。此外,采用图卷积网络和时域卷积网络架构来捕获道路网络的空间和时间信息。为了保护目标城市数据的隐私,我们使用联邦学习来训练目标模型,并提出一种基于空间相似性的参数聚合算法。
Cross-city traffic flow prediction federal transfer learning algorithm based on privacy protection
一种基于隐私保护的跨城市交通流预测联邦迁移学习算法
2024-04-09
Patent
Electronic Resource
Chinese
IPC: | G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G06Q Datenverarbeitungssysteme oder -verfahren, besonders angepasst an verwaltungstechnische, geschäftliche, finanzielle oder betriebswirtschaftliche Zwecke, sowie an geschäftsbezogene Überwachungs- oder Voraussagezwecke , DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Transfer Learning for Cross-City Traffic Prediction to Solve Data Scarcity
Transportation Research Record | 2024
|Federal learning traffic prediction method based on traffic mode classification
European Patent Office | 2025
|Distributed traffic flow prediction method based on personalized federal map signal learning
European Patent Office | 2024
|