Large-scale digital platforms generate billions of timestamped user-item interactions (events) that are crucial for predicting user attributes in, e.g., fraud prevention and recommendations. While self-supervised learning (SSL) effectively models the temporal order of events, it typically overlooks the global structure of the user-item interaction graph. To bridge this gap, we propose three model-agnostic strategies for integrating this structural information into contrastive SSL: enriching event embeddings, aligning client representations with graph embeddings, and adding a structural pretext task. Experiments on four financial and e-commerce datasets demonstrate that our approach consistently improves the accuracy (up to a 2.3% AUC) and reveals that graph density is a key factor in selecting the optimal integration strategy.
@article{arxiv.2604.09085,
title = {Beyond Isolated Clients: Integrating Graph-Based Embeddings into Event Sequence Models},
author = {Harry Proshian and Nikita Severin and Sergey Nikolenko and Kireev Ivan and Andrey Savchenko and Ivan Sergeev and Maria Postnova and Ilya Makarov},
journal= {arXiv preprint arXiv:2604.09085},
year = {2026}
}
Comments
Short paper accepted at ACM Web Conference 2026 (WWW '26)