RobustSTL: A Robust Seasonal-Trend Decomposition Algorithm for Long Time Series

• DOI: 10.1609/aaai.v33i01.33015409
• Publication Date: 2019-08-30T07:32:29Z
• Authors: Qingsong Wen; Jingkun Gao; Xiaomin Song; Liang Sun; Huan Xu; Shenghuo Zhu
• Keywords: Seasonality; Robustness; Seasonal adjustment

Decomposing complex time series into trend, seasonality, and remainder components is an important task to facilitate anomaly detection forecasting. Although numerous methods have been proposed, there are still many characteristics exhibiting in real-world data which not addressed properly, including 1) ability handle seasonality fluctuation shift, abrupt change trend reminder; 2) robustness on with anomalies; 3) applicability long period. In the paper, we propose a novel generic decomposition algorithm address these challenges. Specifically, extract component robustly by solving regression problem using least absolute deviations loss sparse regularization. Based extracted apply non-local seasonal filtering component. This process repeated until accurate obtained. Experiments different synthetic datasets demonstrate that our method outperforms existing solutions.