D-PAD: Deep-Shallow Multi-Frequency Patterns Disentangling for Time Series Forecasting

TL;DR

D-PAD is a novel neural network that disentangles multi-frequency patterns in time series data using a multi-component decomposition and a progressive extraction approach, leading to improved forecasting accuracy.

Contribution

It introduces a deep-shallow multi-frequency disentangling framework with novel decomposition and fusion modules for enhanced time series forecasting.

Findings

Achieves state-of-the-art performance on seven datasets.

Outperforms baselines by 9.48% in MSE and 7.15% in MAE.

Effectively separates frequency components for better modeling.

Abstract

In time series forecasting, effectively disentangling intricate temporal patterns is crucial. While recent works endeavor to combine decomposition techniques with deep learning, multiple frequencies may still be mixed in the decomposed components, e.g., trend and seasonal. Furthermore, frequency domain analysis methods, e.g., Fourier and wavelet transforms, have limitations in resolution in the time domain and adaptability. In this paper, we propose D-PAD, a deep-shallow multi-frequency patterns disentangling neural network for time series forecasting. Specifically, a multi-component decomposing (MCD) block is introduced to decompose the series into components with different frequency ranges, corresponding to the "shallow" aspect. A decomposition-reconstruction-decomposition (D-R-D) module is proposed to progressively extract the information of frequencies mixed in the components, corresponding to the "deep" aspect. After that, an interaction and fusion (IF) module is used to further analyze the components. Extensive experiments on seven real-world datasets demonstrate that D-PAD achieves the state-of-the-art performance, outperforming the best baseline by an average of 9.48% and 7.15% in MSE and MAE, respectively.