Time Series Foundational Models: Their Role in Anomaly Detection and Prediction
It addresses the underexplored application of TSFMs in anomaly detection and prediction, highlighting their limitations for practitioners in time series analysis, but is incremental as it builds on existing concerns about TSFMs.
This paper critically evaluates time series foundational models (TSFM) for anomaly detection and prediction, finding that traditional statistical and deep learning models often match or outperform TSFMs in these tasks, with TSFMs requiring high computational resources and failing to capture sequential dependencies effectively.
Time series foundational models (TSFM) have gained prominence in time series forecasting, promising state-of-the-art performance across various applications. However, their application in anomaly detection and prediction remains underexplored, with growing concerns regarding their black-box nature, lack of interpretability and applicability. This paper critically evaluates the efficacy of TSFM in anomaly detection and prediction tasks. We systematically analyze TSFM across multiple datasets, including those characterized by the absence of discernible patterns, trends and seasonality. Our analysis shows that while TSFMs can be extended for anomaly detection and prediction, traditional statistical and deep learning models often match or outperform TSFM in these tasks. Additionally, TSFMs require high computational resources but fail to capture sequential dependencies effectively or improve performance in few-shot or zero-shot scenarios. \noindent The preprocessed datasets, codes to reproduce the results and supplementary materials are available at https://github.com/smtmnfg/TSFM.