Bayesian Predictive Synthesis for Dynamic Networks: Forecasting and Identifying Structural Mechanisms

2026-06-18

AI Summary

This work addresses the challenge that dominant structural mechanisms—such as communities, geometric embeddings, or hubs—in dynamic networks evolve over time, rendering static or fixed-weight models inadequate for accurate link prediction and mechanism identification. To overcome this limitation, the authors propose a dynamic Bayesian predictive synthesis framework, which treats each mechanism as a predictive agent and adaptively fuses their outputs through a time-varying weighting layer while simultaneously inferring each mechanism’s contribution. By introducing sparse safe parametrization and a theory of mechanism identifiability, the model enables weight estimation from a single observed graph. The framework also establishes an optimal tracking cost for mechanism switching and reduces to calibrated link prediction under a single snapshot. Experiments on both real-world and synthetic data demonstrate that the method achieves highly accurate, probabilistically calibrated edge predictions and reliably recovers the prevailing structural mechanism.

Abstract

Networks are shaped by competing structural mechanisms, such as communities, geometry, or hubs. In a dynamic network the most predictive mechanism can change, and a model tied to one mechanism, or to fixed weights, cannot adapt as the dominant structure shifts. We develop dynamic Bayesian predictive synthesis for networks, in which a mechanism is an agent forecasting the next snapshot's edges and a synthesis layer combines them with time-varying weights. At each step the method returns a calibrated edge forecast and inference on the mechanism weights, with intervals valid given the fitted agents, so it also reports which mechanism is most informative. Inference of this kind requires a sparse-safe parametrization and an identification theory, under which a single graph identifies and estimates the weights. A sharp threshold separates distinguishable from indistinguishable mechanisms, a change in the active mechanism is tracked at an optimal per-switch cost, and for a single snapshot the method reduces to calibrated link prediction. On real networks, simulations, and benchmarks, the synthesis gives accurate, calibrated forecasts and recovers the leading mechanism when

Problem

• dynamic networks
• structural mechanisms
• Bayesian predictive synthesis
• mechanism adaptation
• edge forecasting

Innovation

• Bayesian predictive synthesis
• dynamic networks
• structural mechanisms
• time-varying weights
• mechanism identification