Abstract:The past few years have witnessed the rapid development in multivariate time series forecasting. The key to accurate forecasting results is capturing the long-term dependency between each time step (cross-time dependency) and modeling the complex dependency between each variable (cross-variable dependency) in multivariate time series. However, recent methods mainly focus on the cross-time dependency but seldom consider the cross-variable dependency. To fill this gap, we find that convolution, a traditional technique but recently losing steam in time series forecasting, meets the needs of respectively capturing the cross-time and cross-variable dependency. Based on this finding, we propose a modern pure convolution structure, namely Cross-LKTCN, to better utilize both cross-time and cross-variable dependency for time series forecasting. Specifically in each Cross-LKTCN block, a depth-wise large kernel convolution with large receptive field is proposed to capture cross-time dependency, and then two successive point-wise group convolution feed forward networks are proposed to capture cross-variable dependency. Experimental results on real-world benchmarks show that Cross-LKTCN achieves state-of-the-art forecasting performance and improves the forecasting accuracy significantly compared with existing convolutional-based models and cross-variable methods.

What problem does this paper attempt to address?

The paper aims to address two key issues in multivariate time series forecasting: cross-time dependency and cross-variable dependency. Specifically: 1. **Cross-time dependency**: Refers to the long-term temporal dependencies between each time step. 2. **Cross-variable dependency**: Refers to the complex dependencies between different variables in a multivariate time series. Most existing methods focus on capturing cross-time dependency while paying less attention to cross-variable dependency. To address this shortcoming, the paper proposes a modern pure convolutional structure—Cross-LKTCN (Cross-Large Kernel Temporal Convolution Network), which utilizes traditional convolution techniques to capture both cross-time and cross-variable dependencies separately, thereby improving the accuracy of time series forecasting. ### Main Contributions - Proposed successive point-wise group convolution FFNs, which efficiently capture cross-variable dependencies based on the decoupling properties of group convolution and depth-wise separable convolution. Experimental results show that this method surpasses existing cross-variable methods and achieves state-of-the-art performance, highlighting the importance of cross-variable dependency in time series forecasting. - Inspired by the latest large kernel 2D convolutions in the computer vision field, introduced depth-wise large kernel convolution to expand the receptive field and enhance the modeling capability of cross-time dependencies. Experimental results demonstrate that this method better leverages the potential of convolution in time series forecasting, outperforming other existing convolutional models. - Extensive experiments on 9 real-world benchmark datasets demonstrate the effectiveness of Cross-LKTCN. Compared to existing convolutional models and cross-variable methods, Cross-LKTCN achieves significant relative improvements in multivariate time series forecasting (27.4% and 52.3%, respectively) and reaches state-of-the-art performance levels. Through these contributions, Cross-LKTCN not only improves the accuracy of time series forecasting but also showcases the great potential of convolutional models and cross-variable methods in this field.

Cross-LKTCN: Modern Convolution Utilizing Cross-Variable Dependency for Multivariate Time Series Forecasting Dependency for Multivariate Time Series Forecasting

Causal augmented ConvNet: A temporal memory dilated convolution model for long-sequence time series prediction

ModernTCN: A Modern Pure Convolution Structure for General Time Series Analysis.

Spatial-Temporal Convolutional Transformer Network for Multivariate Time Series Forecasting

TimeCNN: Refining Cross-Variable Interaction on Time Point for Time Series Forecasting

Crossformer: Transformer Utilizing Cross-Dimension Dependency for Multivariate Time Series Forecasting

CrossWaveNet: A dual-channel network with deep cross-decomposition for Long-term Time Series Forecasting

Adaptive Convolutional Forecasting Network Based on Time Series Feature-Driven

LagCNN: A Fast Yet Effective Model for Multivariate Long-term Time Series Forecasting

CVTN: Cross Variable and Temporal Integration for Time Series Forecasting

DDTCN: Decomposed dimension time-domain convolutional neural network along spatial dimensions for multiple long-term series forecasting

TCLN: A Transformer-based Conv-LSTM Network for Multivariate Time Series Forecasting

Biased Temporal Convolution Graph Network for Time Series Forecasting with Missing Values.

Multivariate Time Series Modeling and Forecasting with Parallelized Convolution and Decomposed Sparse-Transformer

Probabilistic Forecasting with Temporal Convolutional Neural Network

Correlating Time Series with Interpretable Convolutional Kernels

ConvTimeNet: A Deep Hierarchical Fully Convolutional Model for Multivariate Time Series Analysis

A Multivariate Time Series Prediction Method Based on Convolution-Residual Gated Recurrent Neural Network and Double-Layer Attention

Improving long-term multivariate time series forecasting with a seasonal-trend decomposition-based 2-dimensional temporal convolution dense network

EffiCANet: Efficient Time Series Forecasting with Convolutional Attention

Client: Cross-variable Linear Integrated Enhanced Transformer for Multivariate Long-Term Time Series Forecasting