Project CEMAPRE internal
| Title | Classification and clustering of time series with data-driven fragmented statistics (cont.) |
| Participants | Jorge Caiado (Principal Investigator), Nuno Crato |
| Summary | This research aims to advance methodologies for classifying and clustering time series in challenging scenarios characterized by noisy data and unknown structures. Specifically, the study introduces innovative discrepancy metrics that leverage significant autocorrelations and partial autocorrelations to calculate distances between time series. Through an extensive simulation study involving diverse linear, stationary, near nonstationary, and nonstationary models, the performance and robustness of the proposed metrics are evaluated. The study also explores the practical applicability of these methods in capturing meaningful patterns in real-world data, particularly in economic and financial time series. By addressing limitations of existing approaches, this project contributes to the development of robust, data-driven tools for analyzing complex time series datasets. |