F2STNet: Fair and Federated Spectral-Temporal Modeling for Graph Forecasting

2026-08-09

AI Summary

This work addresses the challenges of collaborative modeling and insufficient prediction fairness for spatiotemporal graph-structured data in decentralized heterogeneous environments by proposing the F²STNet framework. The method integrates truncated graph Fourier features, a lightweight diagonal state-space temporal encoder, and graph convolutional networks, and introduces a federated aggregation mechanism—Fairness-aware Federated Aggregation (FFA)—based on validation loss and fair scheduling to enable joint spectral-temporal modeling and equitable optimization across clients. Experimental results on the PeMS04, HZMetro, and KnowAir datasets demonstrate that F²STNet not only enhances overall prediction accuracy but also significantly improves the performance of the worst-performing clients while reducing performance disparity among clients.

Abstract

Spatiotemporal prediction on graph-structured data is central to traffic forecasting and environmental monitoring, yet decentralized and heterogeneous data complicate both sequence modeling and collaborative training. We propose F$^2$STNet, a federated forecasting framework that combines truncated graph-Fourier features, a lightweight diagonal state-space temporal encoder, graph convolution, and Fairness-aware Federated Aggregation (FFA). The spectral branch exposes graph-frequency structure, while the state-space layer models long temporal dependencies with linear complexity in the sequence length. FFA adjusts the FedAvg prior using client validation losses and an increasing fairness schedule. Experiments on PeMS04, HZMetro, and KnowAir show favorable forecasting accuracy relative to the evaluated baselines; federated experiments on PeMS04 additionally improve worst-client and client-dispersion metrics.

Problem

• spatiotemporal prediction
• graph-structured data
• decentralized data
• heterogeneous data
• federated learning

Innovation

• Federated Learning
• Graph Forecasting
• State-Space Model
• Fairness-aware Aggregation
• Spectral-Temporal Modeling