HiSTM: Hierarchical Spatiotemporal Mamba for Cellular Traffic Forecasting

2025-08-07

AI Summary

Cellular traffic forecasting is highly challenging due to complex spatiotemporal dependencies induced by user mobility. To address this, we propose an efficient and accurate deep learning model featuring a dual-branch spatial encoder that captures multi-granularity spatial correlations, integrated with a Mamba-based temporal module leveraging selective state space modeling to explicitly encode long-range temporal dynamics and multi-scale spatiotemporal couplings. Evaluated on real-world datasets, our model reduces MAE by 29.4% compared to the STN baseline while decreasing parameter count by 94%, significantly improving both long-horizon prediction accuracy and generalization capability. The core contribution lies in pioneering the integration of lightweight selective state space modeling into cellular traffic forecasting—achieving, for the first time, a unified balance of high accuracy, ultra-low parameter overhead, and strong robustness.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Mixture of Experts (MoE)
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Cellular traffic forecasting is essential for network planning, resource allocation, or load-balancing traffic across cells. However, accurate forecasting is difficult due to intricate spatial and temporal patterns that exist due to the mobility of users. Existing AI-based traffic forecasting models often trade-off accuracy and computational efficiency. We present Hierarchical SpatioTemporal Mamba (HiSTM), which combines a dual spatial encoder with a Mamba-based temporal module and attention mechanism. HiSTM employs selective state space methods to capture spatial and temporal patterns in network traffic. In our evaluation, we use a real-world dataset to compare HiSTM against several baselines, showing a 29.4% MAE improvement over the STN baseline while using 94% fewer parameters. We show that the HiSTM generalizes well across different datasets and improves in accuracy over longer time-horizons.

Problem

• Accurate cellular traffic forecasting with complex patterns
• Balancing accuracy and computational efficiency in AI models
• Capturing hierarchical spatiotemporal dependencies in network traffic

Innovation

• Combines dual spatial encoder with Mamba-based temporal module
• Uses selective state space methods for pattern capture
• Achieves high accuracy with significantly fewer parameters