Beyond Isolated Clients: Integrating Graph-Based Embeddings into Event Sequence Models

Harry Proshian; Nikita Severin; Sergey Nikolenko; Kireev Ivan; Andrey Savchenko; Ivan Sergeev; Maria Postnova; Ilya Makarov

cs.LGApr 10, 2026

Beyond Isolated Clients: Integrating Graph-Based Embeddings into Event Sequence Models

Authors: Harry Proshian, Nikita Severin, Sergey Nikolenko, Kireev Ivan, Andrey Savchenko, Ivan Sergeev, Maria Postnova, Ilya Makarov

Organizations: Steklov Institute of Mathematics Saint Petersburg, Russia · Independent Researcher Belgrade, Serbia · ISP RAS Steklov Institute of Mathematics Saint Petersburg, Russia · Sber AI Lab Moscow, Russia ·

Abstract

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.

Figures & tables

Figure 1. Global interaction graph from event sequences

	Gender	Age	MTS-ML-Cup	Internal
Client entities	15 K	50 K	415 K	1M
Non-client entities	184	204	197 K	418
Interactions	6.85 M	44 M	323 M	530M
Labeled clients	8.4 K	30 K	270 K	1M
training set	7.56 K	27 K	243 K	900K
test set	840	3000	27K	100K

Table 1. Dataset statistics.

Method		Gender		Age	Internal
Base	Ours	AUC	Acc	Acc	AUC	Acc
Popular class		0.558	0.558	0.252	0.333	0.333
CoLES		0.877	0.793	0.637	0.905	0.748
CPC		0.851	0.768	0.602	0.899	0.741
BT		0.865	0.779	0.634	0.895	0.734
Tabformer		0.847	0.764	0.601	0.894	0.733

Table 2. Best experimental results.

Approach	AUC	Acc
Base CoLES	0.877	0.793
+ Reg	0.855	0.768
+ Emb SAGE , joint training:
with GNN loss	0.881	0.799
without GNN loss	0.872	0.789
+ Emb SAGE , seq. training	0.886	0.808

Table 3. Graph-based features for CoLES on Gender

Figure 2. Latent space dissimilarity scores; GrW/GrUnw - features from the weighted/unweighted adjacency matrix.

Self-Supervised Learning

Self-supervised learning uses structure within unlabeled data to train representations and models for downstream tasks.

See all 249 papers

Embedding-Aware Feature Discovery: Bridging Latent Representations and Interpretable Features in Event Sequences
Mar 16, 2026Artem Sakhno, Ivan Sergeev, Alexey Shestov +5Large Language Model-Based Feature Engineering Automated Feature Engineering
Sber AI Lab · ISP RAS Research Center for Trusted AI
A Foundation Model for Multimodal Event Sequences in Financial Applications
Jul 10, 2026Nikita Rusakov, Vladislav Meshkov, Konstantin Zorin +4Financial Services Multimodal Pretraining
Sber Moscow, Russian Federation · Sber AI Lab Moscow, Russian Federation
User-Centric Modeling of Transactional Sequences with Explainable State Space Models
Jul 22, 2026Ivan PalaginSelective SSMs User Modeling
HSE University Moscow, Russia

Mar 16, 2026J/K move · Enter open · S save

Embedding-Aware Feature Discovery: Bridging Latent Representations and Interpretable Features in Event Sequences

Artem Sakhno, Ivan Sergeev, Alexey Shestov, Omar Zoloev+4

Sber AI Lab · ISP RAS Research Center for Trusted AI

Large Language Model-Based Feature Engineering Automated Feature Engineering Sequence Modeling Predictive Representation Learning

Open paper AskPDF arXiv

Industrial financial systems operate on temporal event sequences such as transactions, user actions, and system logs. While recent research emphasizes representation learning and large language models, production systems continue to rely heavily on handcrafted statistical features due to their interpretability, robustness under limited supervision, and strict latency constraints. This creates a persistent disconnect between learned embeddings and feature-based pipelines. We introduce Embedding-Aware Feature Discovery (EAFD), a unified framework that bridges this gap by coupling pretrained event-sequence embeddings with a self-reflective LLM-driven feature generation agent. EAFD iteratively discovers, evaluates, and refines features directly from raw event sequences using two complementary criteria: alignment, which explains information already encoded in embeddings, and complementarity, which identifies predictive signals missing from them. Across both open-source and industrial transaction benchmarks, EAFD consistently outperforms embedding-only and feature-based baselines, achieving relative gains of up to

+5.8\%

over state-of-the-art pretrained embeddings, resulting in new state-of-the-art performance across event-sequence datasets.

Abstract

Self-Supervised Learning

Explore similar work

Embedding-Aware Feature Discovery: Bridging Latent Representations and Interpretable Features in Event Sequences

A Foundation Model for Multimodal Event Sequences in Financial Applications

User-Centric Modeling of Transactional Sequences with Explainable State Space Models