Financial fraud detection improved by role-aware graph learning with LLMs
LLM-Enhanced Heterogeneous Graph Learning with Business Semantics for Financial Fraud Detection
Artificial Intelligence
Summary
Detecting financial fraud is hard because different actors in transactions, like cards and merchants, have different normal behaviors. The authors created a method that uses large language models (LLMs) to better understand the business meaning of transaction data by considering the roles of these actors. Their approach turns structured transaction details into richer meanings and uses this to find suspicious activities in transaction networks. Tests on real data show their method works better than existing ones at spotting fraud.
What this means in practice
- •For financial fraud teams: Improve transaction monitoring systems by explicitly modeling business roles to better detect suspicious patterns in card and merchant activities.
- •For payment platform developers: Build fraud detection features that leverage structured transaction data enriched with semantic role-based insights from large language models.
Authors
Yixuan Chen, Hongyu Zhan, Jie Sheng, Weiyu Han, Shuai Chen, Tianyi Zhang, Xiao Tan, Jun Xia
Abstract
Financial fraud detection increasingly relies on heterogeneous transaction graphs, in which entities such as cards and merchants play distinct business roles and the risk meaning of an attribute depends on the entity's role: a high transaction volume may be routine for a merchant but unusual for a card. Heterogeneous graph neural networks preserve typed entities and relations but typically treat these attributes as role-agnostic inputs and leave their business meaning implicit. Meanwhile, most existing methods integrating large language models (LLMs) with graphs focus on homogeneous graphs or heterogeneous graphs with naturally available textual attributes, leaving their use on transaction graphs with structured attributes underexplored. To address this issue, we propose SemHGT, an LLM-enhanced framework that interprets structured attributes according to entity roles and makes the business semantics explicit in three stages: (1) role-aware prompts elicit risk-related business descriptions from node attributes, and a text encoder maps them to semantic vectors that enrich node representations; (2) semantic similarity guides multi-round sampling of target-specific path instances under transaction-to-transaction schemas; (3) schema-level tokens that combine local path evidence with global neighbor summaries are processed together with the target transaction by a unified transformer. Across three benchmark datasets, SemHGT leads the evaluated baselines on most dataset--metric comparisons. Further analyses provide additional evidence for the effectiveness of SemHGT's design and clarify how its semantic modeling and evidence selection contribute to fraud detection. Code is available at https://anonymous.4open.science/r/SemHGT-528A/.