A Unified Replay-based Continuous Learning Framework for Spatio-Temporal Prediction on Streaming Data

TL;DR

The paper tackles prediction on streaming spatio-temporal data by introducing URCL, a unified replay-based continuous learning framework. It combines a replay buffer with a spatio-temporal mixup (STMixup) and a holistic representation module (STSimSiam) to preserve historical knowledge while training on new data, and it uses a spatio-temporal autoencoder with a shared encoder for prediction. Empirical results on four urban datasets show URCL achieves top-tier accuracy and robustness to concept drift, with ablations confirming the value of each component and the approach’s backbone flexibility. The work offers a practical, scalable solution for real-time spatio-temporal forecasting in dynamic environments, with potential impact on traffic, mobility, and environmental sensing applications.

Abstract

The widespread deployment of wireless and mobile devices results in a proliferation of spatio-temporal data that is used in applications, e.g., traffic prediction, human mobility mining, and air quality prediction, where spatio-temporal prediction is often essential to enable safety, predictability, or reliability. Many recent proposals that target deep learning for spatio-temporal prediction suffer from so-called catastrophic forgetting, where previously learned knowledge is entirely forgotten when new data arrives. Such proposals may experience deteriorating prediction performance when applied in settings where data streams into the system. To enable spatio-temporal prediction on streaming data, we propose a unified replay-based continuous learning framework. The framework includes a replay buffer of previously learned samples that are fused with training data using a spatio-temporal mixup mechanism in order to preserve historical knowledge effectively, thus avoiding catastrophic forgetting. To enable holistic representation preservation, the framework also integrates a general spatio-temporal autoencoder with a carefully designed spatio-temporal simple siamese (STSimSiam) network that aims to ensure prediction accuracy and avoid holistic feature loss by means of mutual information maximization. The framework further encompasses five spatio-temporal data augmentation methods to enhance the performance of STSimSiam. Extensive experiments on real data offer insight into the effectiveness of the proposed framework.

Figures (8)

• Figure 1: URCL Framework Overview
• Figure 2: Spatio-temporal Data Augmentation
• Figure 3: Illustation of STEncoder
• Figure 4: Illustation of STDecoder
• Figure 5: Illustration of Replay-based Streaming Spatio-Temporal Data Prediction

Theorems & Definitions (3)

• Definition 1: Sensor Network
• Definition 2: Spatio-temporal Observation
• Definition 3: Streaming Spatio-temporal Data Sequence