Predicting Graph Structure via Adapted Flux Balance Analysis

2025-07-08

AI Summary

Traditional dynamic graph forecasting methods assume a fixed node set, limiting their ability to model node arrivals and departures. To address this, we propose a graph evolution prediction framework that jointly leverages time-series modeling and adaptive Flow Balance Analysis (FBA). Our approach relaxes the static-node assumption by integrating linear programming optimization with a preferential attachment prior, enabling joint learning of node inflow/outflow dynamics and edge evolution patterns—thus supporting topological growth and structural reconfiguration. Experiments on real-world dynamic networks (UCI Message, Facebook, Bitcoin) and synthetic benchmarks demonstrate significant improvements: average AUC for future-edge prediction increases by 8.2%, and node activity inference accuracy is notably enhanced. The framework improves both generalizability and interpretability of dynamic graph forecasting, offering principled insights into evolutionary mechanisms through flow-balanced dynamics.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

Many dynamic processes such as telecommunication and transport networks can be described through discrete time series of graphs. Modelling the dynamics of such time series enables prediction of graph structure at future time steps, which can be used in applications such as detection of anomalies. Existing approaches for graph prediction have limitations such as assuming that the vertices do not to change between consecutive graphs. To address this, we propose to exploit time series prediction methods in combination with an adapted form of flux balance analysis (FBA), a linear programming method originating from biochemistry. FBA is adapted to incorporate various constraints applicable to the scenario of growing graphs. Empirical evaluations on synthetic datasets (constructed via Preferential Attachment model) and real datasets (UCI Message, HePH, Facebook, Bitcoin) demonstrate the efficacy of the proposed approach.

Problem

• Predict future graph structure in dynamic networks
• Overcome vertex immutability assumption in existing methods
• Adapt flux balance analysis for growing graph constraints

Innovation

• Combines time series prediction with adapted FBA
• Adapts FBA for growing graph constraints
• Validated on synthetic and real datasets