Attention-Guided Deep Adversarial Temporal Subspace Clustering (A-DATSC) Model for multivariate spatiotemporal data

TL;DR

The paper addresses unsupervised clustering of 4D multivariate spatiotemporal data by framing it as a deep subspace clustering problem on nonlinear manifolds. It introduces A-DATSC, a generator–discriminator framework that combines a ConvLSTM-based spatiotemporal encoder, a Bidirectional Temporal Graph Attention self-expressive network, a DEC-style clustering head with temperature and balancing, a self-expressive temporal layer, and an energy-based subspace discriminator. Key contributions include a parameter-efficient, structure-preserving generator, a graph-attention self-expressive network capturing local/global temporal-spatial dependencies, and an energy-based discriminator enforcing subspace geometry, with strong empirical results on three real-world datasets. The approach offers robust unsupervised clustering of complex spatiotemporal phenomena, with potential impact on climate science, epidemiology, and resource planning due to improved regime segmentation and trajectory understanding.

Abstract

Deep subspace clustering models are vital for applications such as snowmelt detection, sea ice tracking, crop health monitoring, infectious disease modeling, network load prediction, and land-use planning, where multivariate spatiotemporal data exhibit complex temporal dependencies and reside on multiple nonlinear manifolds beyond the capability of traditional clustering methods. These models project data into a latent space where samples lie in linear subspaces and exploit the self-expressiveness property to uncover intrinsic relationships. Despite their success, existing methods face major limitations: they use shallow autoencoders that ignore clustering errors, emphasize global features while neglecting local structure, fail to model long-range dependencies and positional information, and are rarely applied to 4D spatiotemporal data. To address these issues, we propose A-DATSC (Attention-Guided Deep Adversarial Temporal Subspace Clustering), a model combining a deep subspace clustering generator and a quality-verifying discriminator. The generator, inspired by U-Net, preserves spatial and temporal integrity through stacked TimeDistributed ConvLSTM2D layers, reducing parameters and enhancing generalization. A graph attention transformer based self-expressive network captures local spatial relationships, global dependencies, and both short- and long-range correlations. Experiments on three real-world multivariate spatiotemporal datasets show that A-DATSC achieves substantially superior clustering performance compared to state-of-the-art deep subspace clustering models.

Figures (1)

• Figure 1: Architecture of Attention-guided deep adversarial temporal subspace clustering (A-DATSC) model