Dynamic Feature Extraction and Prediction for High Dimensional Time Series with Seasonality
Bingyuan Li, Yang Wang, Yang Zhao, Lishuai Li, Yining Dong · 2024
In this paper, a novel reduced-dimensional seasonal autoregressive modeling algorithm with a canonical correlation analysis objective (RDSAR-CCA) is developed for dynamic feature extraction and prediction in high-dimensional time series with seasonality. The proposed algorithm estimates a seasonal reduced-dimensional dynamic model for extracting and modeling the latent dynamics within the data. This approach facilitates dynamic latent variable (DLV) analysis in high-dimensional seasonal time series. The DLVs extracted by the proposed algorithm are orthogonal and ranked in descending order of predictability, which simplifies interpretation and enhances visualization. The effectiveness and superiority of the proposed RDSAR-CCA algorithm are evaluated on the passenger flow data from the Shenzhen metro system.