BjTT is a significant multimodal dataset for traffic prediction, offering over 32,000 time-series traffic records with text descriptions of events. It aims to address the limitations of existing datasets by providing diverse data types and new data modalities. The dataset includes 1,260 road vertices and 32,400 time steps, spanning three months of data collection within Beijing's fifth ring expressway area. By incorporating event descriptions with traffic data, BjTT enables researchers to tackle more challenging and practical problems in the field of traffic prediction.
翻譯成其他語言
從原文內容
arxiv.org
深入探究