With the construction and development of highway informatization, highway traffic has entered the era of big data. The reliability and validity of data are the basis and prerequisite to ensure the reliable operation for the highway information systems. However, data loss is inevitable due to equipment failure and external factors in the process of data collection. It is important to study the imputation method of traffic data. This paper proposes a traffic flow data imputation method based on Feature Fusion Attentional Imputation Network (FFAI-Net). Traffic data are input as a matrix. After shallow feature extraction, the attention mechanism is used to extract the weights of different features for traffic flow data. The feature fusion model gives adaptive learning weights to different levels of features, and establishes a mapping relationship between missing data and real data. With California highway data PeMS03 as experimental data, the test results show that the proposed FFAI-Net has high accuracy at different missing rates.
Traffic Flow Data Imputation Based on Feature Fusion Attention Imputation Network
2023-10-28
1902756 byte
Conference paper
Electronic Resource
English
Missing Traffic Data Imputation based on Tensor Completion and Graph Network Fusion
Transportation Research Record | 2025
|Spatial Network-Wide Traffic Flow Imputation with Graph Neural Network
Springer Verlag | 2025
|