CC-Time achieves state-of-the-art prediction accuracy in time series modeling, indicating improved integration of language models and time series attributes.
Key Points
CC-Time achieves state-of-the-art predictive accuracy across various real-world datasets, showcasing its effectiveness.
The approach combines temporal dependency modeling and channel correlations from both time series and corresponding text descriptions.
Using cross-model fusion, CC-Time integrates knowledge from pre-trained language models and traditional time series models.
Extensive experiments demonstrate improvements in prediction accuracy in both full-data training and few-shot learning scenarios.