Distributed Lag Transformer based on Time-Variable-Aware Learning for Explainable Multivariate Time Series Forecasting

TL;DR

DLFormer is a novel Transformer model designed for multivariate time series forecasting that combines accuracy with interpretability by modeling temporal dependencies and variable influences explicitly.

Contribution

The paper introduces DLFormer, a Transformer architecture that incorporates distributed lag embedding and time-variable-aware learning for explainable and scalable multivariate time series forecasting.

Findings

Achieves state-of-the-art accuracy on benchmark datasets.

Provides robust, interpretable insights into variable and temporal dynamics.

Bridges the gap between forecasting performance and explainability.

Abstract

Time series data is a key element of big data analytics, commonly found in domains such as finance, healthcare, climate forecasting, and transportation. In large scale real world settings, such data is often high dimensional and multivariate, requiring advanced forecasting methods that are both accurate and interpretable. Although Transformer based models perform well in multivariate time series forecasting (MTSF), their lack of explainability limits their use in critical applications. To overcome this, we propose Distributed Lag Transformer (DLFormer), a novel Transformer architecture for explainable and scalable MTSF. DLFormer integrates a distributed lag embedding and a time variable aware learning (TVAL) mechanism to structurally model both local and global temporal dependencies and explicitly capture the influence of past variables on future outcomes. Experiments on ten benchmark and real world datasets show that DLFormer achieves state of the art predictive accuracy while offering robust, interpretable insights into variable wise and temporal dynamics. These results highlight ability of DLFormer to bridge the gap between performance and explainability, making it highly suitable for practical big data forecasting tasks.