---
title: "CASE-Net: Deep Spatio-Temporal Representation Learning via Causal Attention and Channel Recalibration for Multivariate Time Series Classification"
authors:
  - "Fan Zhang"
  - "Yating Cui"
  - "Hua Wang"
doi: "10.24963/ijcai.2026/354"
url: "https://ijcai.org/proceedings/2026/354"
pdf: "https://ijcai.org/proceedings/2026/0354.pdf"
venue: "Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence"
year: 2026
pages: "3187-3195"
track: "Main Track"
keywords:
  - "Machine Learning: Deep learning architectures"
  - "Machine Learning: Time series and data streams"
  - "Data Mining: Mining spatial and/or temporal data"
---

# CASE-Net: Deep Spatio-Temporal Representation Learning via Causal Attention and Channel Recalibration for Multivariate Time Series Classification

Fan Zhang, Yating Cui, Hua Wang

*Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence*, pages 3187-3195. DOI: [10.24963/ijcai.2026/354](https://doi.org/10.24963/ijcai.2026/354)

## Abstract

Multivariate time series (MTS) classification is foundational to pervasive computing and financial analysis, yet existing multi-scale paradigms are often constrained by suboptimal representation fidelity. We identify two critical bottlenecks: temporal non-causality in standard encoders that induces temporal confounding in non-stationary dynamics, and the absence of explicit channel saliency mechanisms that allows noise to contaminate the latent space. To address these challenges, we propose the Causal Attention and Spatio-temporal Encoder Network (CASE-Net), an architecture designed for structural manifold pre-conditioning. CASE-Net synergizes a Causal Temporal Encoder, which enforces physical arrow-of-time constraints via masked self-attention and causal convolutions, with an Adaptive Channel Recalibration module functioning as an information bottleneck to suppress detrimental noise. Comprehensive evaluations across six heterogeneous domains demonstrate that CASE-Net establishes new state-of-the-art benchmarks on four tasks, achieving a peak accuracy of 98.6% on the AWR dataset and superior robustness in non-stationary regimes.