Abstract:Time series forecasting has been playing an important role in decision making, control, and monitoring across various fields. Specifically, the forecasting of nonstationarity time series remains a challenging problem where traditional time series modeling may not fully capture temporal dynamics. Recent studies of applying machine learning (ML) or more advanced hybrid models combining the ML and decomposition methods have shown their flexible nonstationary and nonlinear modeling capability. However, the end-effect problem introduced by the decomposition methods still introduces significant forecasting errors because of the unknown realm beyond the time series boundary. Therefore, a novel method applying a decomposition-guided mechanism is proposed in this work to eliminate the end effect problem while inheriting the knowledge learned from the decomposition state space to improve the prediction accuracy of such hybrid models in time series forecasting. Additionally, a domain adaptation model is integrated with the proposed mechanism to transfer knowledge from the source domain to the target domain regarding the decomposition state space. In this work, the intrinsic time-scale decomposition and Gaussian process are considered as examples of decomposition and ML methods to demonstrate the proposed mechanism’s effectiveness. Both simulation experiments and real-world case studies are conducted to show that a hybrid model with the proposed mechanism outperforms the conventional time series forecasting model, pure ML, and the original hybrid model in terms of prediction accuracy.

What problem does this paper attempt to address?

This paper attempts to address the challenges in non - stationary time - series forecasting, especially the problem that traditional time - series modeling methods may not fully capture the time dynamics. Specifically, the paper focuses on how to utilize the advantages of the combination of decomposition methods and machine - learning (ML) models for non - stationary time - series forecasting while overcoming the end - effect problem introduced by decomposition methods. The end - effect can lead to a significant increase in prediction error because the decomposition method lacks information about extreme points outside the boundaries of the time series, which causes bias at the two boundaries of the decomposed components. This bias accumulates as the number of components decomposed from the original time series increases. To meet this challenge, the paper proposes a new decomposition - guided mechanism, aiming to eliminate the end - effect problem while inheriting the knowledge learned from the decomposition state space to improve the prediction accuracy of such hybrid models in time - series forecasting. In addition, the paper also integrates a domain adaptation model to transfer knowledge from the source domain to the target domain, especially when dealing with the decomposition state space. Through this method, the paper attempts to provide a more effective high - order non - stationary time - series prediction model. The main contributions of the paper include: 1. Proposing a decomposition - guided mechanism that combines the advantages of decomposition methods and machine - learning methods for non - stationary time - series forecasting while avoiding the end - effect problem in hybrid models. 2. Integrating a domain adaptation model to identify the patterns of target points in the decomposition state space by re - weighting the contributions of source - domain points and boundary points in prediction. Through these methods, the paper demonstrates the effectiveness of the proposed framework in simulation experiments and actual case studies, proving that compared with traditional time - series prediction models, pure machine - learning models, and the original hybrid models, this model has significant advantages in prediction accuracy.

A decomposition-guided mechanism for nonstationary time series forecasting

Complexity-aided time series modeling and forecasting under a decomposition-aggregation framework

A novel decomposed-ensemble time series forecasting framework: capturing underlying volatility information

Enhancing Time Series Forecasting: A Hierarchical Transformer with Probabilistic Decomposition Representation

Multivariate Time Series Forecasting based on Time-Frequency Decomposition

A Hybrid Model for Time Series Prediction Using Adaptive Variational Mode Decomposition

A New Unified Deep Learning Approach with Decomposition-Reconstruction-Ensemble Framework for Time Series Forecasting

Revitalizing Multivariate Time Series Forecasting: Learnable Decomposition with Inter-Series Dependencies and Intra-Series Variations Modeling

Effective LSTMs with Seasonal-Trend Decomposition and Adaptive Learning and Niching-Based Backtracking Search Algorithm for Time Series Forecasting

A novel method for time series prediction based on error decomposition and nonlinear combination of forecasters

A novel general-purpose hybrid model for time series forecasting

A Decomposition-Based Memetic Neural Architecture Search Algorithm for Univariate Time Series Forecasting.

Time series forecasting model for non-stationary series pattern extraction using deep learning and GARCH modeling

TDG4MSF: A temporal decomposition enhanced graph neural network for multivariate time series forecasting

Improving the Transferability of Time Series Forecasting with Decomposition Adaptation

Improving Accuracy Without Losing Interpretability: A ML Approach for Time Series Forecasting

A Hybrid Method with Adaptive Sub-Series Clustering and Attention-Based Stacked Residual LSTMs for Multivariate Time Series Forecasting

A Novel Time Series Forecasting Approach with Multi-Level Data Decomposing and Modeling

Multi-Scale Information Granule-Based Time Series Forecasting Model with Two-Stage Prediction Mechanism

A Hybrid Modelling Method for Time Series Forecasting Based on a Linear Regression Model and Deep Learning

Hybrid deep learning and empirical mode decomposition model for time series applications.