A Robust and Efficient Multi-Scale Seasonal-Trend Decomposition

  title={A Robust and Efficient Multi-Scale Seasonal-Trend Decomposition},
  author={Linxiao Yang and Qingsong Wen and Bo Yang and Liang Sun},
  journal={ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  • Linxiao Yang, Qingsong Wen, +1 author Liang Sun
  • Published 6 June 2021
  • Computer Science, Mathematics, Engineering
  • ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Many real-world time series exhibit multiple seasonality with different lengths. The removal of seasonal components is crucial in numerous applications of time series, including forecasting and anomaly detection. However, many seasonal-trend decomposition algorithms suffer from high computational cost and require a large amount of data when multiple seasonal components exist, especially when the periodic length is long. In this paper, we propose a general and efficient multi-scale seasonal… 

Figures and Tables from this paper

CloudRCA: A Root Cause Analysis Framework for Cloud Computing Platforms
A root cause analysis framework called CloudRCA which makes use of heterogeneous multi-source data including Key Performance Indicators, logs, as well as topology, and extracts important features via state-of-the-art anomaly detection and log analysis techniques and consistently outperforms existing approaches in f1-score across different cloud systems.
RobustPeriod: Robust Time-Frequency Mining for Multiple Periodicity Detection
A robust and general framework for multiple periodicity detection by applying maximal overlap discrete wavelet transform to transform the time series into multiple temporal-frequency scales such that different periodic components can be isolated.


RobustSTL: A Robust Seasonal-Trend Decomposition Algorithm for Long Time Series
This work proposes a novel and generic time series decomposition algorithm that extracts the trend component robustly by solving a regression problem using the least absolute deviations loss with sparse regularization and applies the non-local seasonal filtering to extract the seasonality component.
Fast RobustSTL: Efficient and Robust Seasonal-Trend Decomposition for Time Series with Complex Patterns
This paper extends RobustSTL to handle multiple seasonality and proposes a special generalized ADMM algorithm to perform the decomposition efficiently, and proves that the proposed algorithm converges approximately as standard ADMM while reducing the complexity from O(n2) to O(N log N) for each iteration.
Forecasting Time Series With Complex Seasonal Patterns Using Exponential Smoothing
An innovations state space modeling framework is introduced for forecasting complex seasonal time series such as those with multiple seasonal periods, high-frequency seasonality, non-integer
STR: A Seasonal-Trend Decomposition Procedure Based on Regression
We propose new generic methods for decomposing seasonal data: STR (a Seasonal-Trend decomposition procedure based on Regression) and Robust STR. In some ways, STR is similar to Ridge Regression and
Forecasting Time Series with Multiple Seasonal Patterns
A new approach is proposed for forecasting a time series with multiple seasonal patterns. A state space model is developed for the series using the innovation approach which enables us to develop
Optimal multi-scale patterns in time series streams
This work introduces a method to discover optimal local patterns, which concisely describe the main trends in a time series and introduces a criterion to select the best window sizes, which most concisely capture the key oscillatory as well as aperiodic trends.
Forecasting Time Series With VARMA Recursions on Graphs
It is found that by projecting the data to the graph spectral domain the multivariate model estimation reduces to that of fitting a number of uncorrelated univariate autoregressive–moving-average models and an optimal low-rank data representation can be exploited so as to further reduce the estimation costs.
Forecasting monthly and quarterly time series using STL decomposition
The results suggest that the new forecasting method can perform well relative to four other standard statistical techniques from the literature, namely the ARIMA, Theta, Holt-Winters' and Holt's Damped Trend methods.
Time Series Data Mining
  • Nuno Castro
  • Computer Science
    Encyclopedia of Database Systems
  • 2009
This work focuses particularly on the area of time series motif discovery (Lin and Keogh 2002) , also known as the extraction of recurrent patterns, which are relevant because they summarise the time series of a domain and help the domain expert understand the database at hand.
On Robust Variance Filtering and Change of Variance Detection
This paper adopts Huber loss to suppress outliers both in trend removal and variance filtering, utilize sparse regu-larizations to capture trend and variance changes, and obtain accurate change points locations by using breakpoint detection for centered cumulative sum of the estimated variance.