Adaptive Convolutional Forecasting Network Based on Time Series Feature-Driven

TL;DR

ACNet tackles nonlinear, noisy time series forecasting by combining multi-resolution dilated convolutions for local pattern capture with adaptive pooling for global trends and gated deformable convolutions to learn nonlinear inter-variable relations. It integrates data normalization and wavelet denoising, a temporal-nonlinear feature acquisition module, and a dynamic prediction stage based on the Moore-Penrose pseudo-inverse for fast updates, enabling online recalibration when performance drifts. On 12 real-world datasets, ACNet significantly surpasses state-of-the-art ConvTimeNet and Crossformer baselines while maintaining favorable efficiency, highlighting the importance of adaptive, cross-resolution feature learning for both short- and long-term forecasting. The work demonstrates that combining local/global temporal features with nonlinear adaptive sampling yields robust, scalable TSF with practical applicability across energy, finance, medical, and traffic domains.

Abstract

Time series data in real-world scenarios contain a substantial amount of nonlinear information, which significantly interferes with the training process of models, leading to decreased prediction performance. Therefore, during the time series forecasting process, extracting the local and global time series patterns and understanding the potential nonlinear features among different time observations are highly significant. To address this challenge, we introduce multi-resolution convolution and deformable convolution operations. By enlarging the receptive field using convolution kernels with different dilation factors to capture temporal correlation information at different resolutions, and adaptively adjusting the sampling positions through additional offset vectors, we enhance the network's ability to capture potential nonlinear features among time observations. Building upon this, we propose ACNet, an adaptive convolutional network designed to effectively model the local and global temporal dependencies and the nonlinear features between observations in multivariate time series. Specifically, by extracting and fusing time series features at different resolutions, we capture both local contextual information and global patterns in the time series. The designed nonlinear feature adaptive extraction module captures the nonlinear features among different time observations in the time series. We evaluated the performance of ACNet across twelve real-world datasets. The results indicate that ACNet consistently achieves state-of-the-art performance in both short-term and long-term forecasting tasks with favorable runtime efficiency.

Figures (8)

• Figure 1: Nonlinear Features of Data.
• Figure 2: Multiple convolution schemes.
• Figure 3: Framework of the ACNet.
• Figure 4: Temporal-Nonlinear feature extraction module.
• Figure 5: Visualize the ETTh1 dataset, where the first 96 data points represent the input length.