Link Analysis of Conflict through Signed Spectral Embedding: The Russo-Ukrainian War

In military intelligence, link analysis is a tool used to review relationships between individuals or groups for insight. The difficulty of accurately depicting complex relationships in link diagrams increases as the number of actors and their activities grows. To address this challenge, we explore the use of spectral embedding as an automated approach, which is capable of preserving the intricate details when presenting relationships as graphs. While spectral embeddings are commonly employed to understand implicit structures in social networks, they are conventionally used in networks with single-weighted edges. We build upon our previously developed spectral embedding technique that allow signed edges to represent both negative and positive relationships. Once networks are represented in lower-dimensional spaces, they can be presented as link diagrams, where the geometric proximity between nodes approximates their similarity. We demonstrate the application of this intelligence tool by automatically generating armed-conflict link diagrams in a study of the Russo-Ukrainian War. By placing each entity in an embedded space, this technique enables the detection of clusters, provides quantifiable assessments, and offers a visual representation of the conflict dynamics. This work showcases the potential of using spectral embedding to enhance our understanding of real-life complex social networks, contributing to more informed decision-making in conflict analysis and management.