Towards Consumer Loan Fraud Detection: Graph Neural Networks with Role-Constrained Conditional Random Field

Abstract

Consumer loans, i.e., loans to finance consumers to buy certain types of expenditures, is increasingly popular in e-commerce platform. Different from traditional loans with mortgage, online consumer loans only take personal credit as collateral for loans. Consequently, loan fraud detection is particularly critical for lenders to avoid economic loss. Previous methods mainly leverage applicant's attributes and historical behavior for loan fraud detection. Although these methods gain success at detecting potential charge-offs, yet they perform worse when multiple persons with various roles (e.g., sellers, intermediaries) collude to apply fraudulent loan. To combat this challenge, we consider the problem of loan fraud detection via exploiting roles of users and multi-type social relationships among users. We propose a novel Graph neural network with a Role-constrained Conditional random field, namely GRC, to learn the representation of applicants and detect loan fraud based on the learned representation. The proposed model characterizes the multiple types of relationships via self-attention mechanism and employs conditional random field to constrain users with the same role to have similar representation. We validate the proposed model through experiments in large-scale auto-loan scenario. Extensive experiments demonstrate that our model achieves state-of-the-art results in loan fraud detection on Alipay, one online credit payment service serving more than 450 million users in China.

FraudAuditor: A Visual Analytics Approach for Collusive Fraud in Health Insurance

“…ML methods typically use GNN to detect fraud, as it is powerful for learning a deep representation of nodes. Previous studies are either conducted on homogeneous [6], [14] or heterogeneous graphs [7], [8], [15]. Wang et al [6] constructed a network of reviewers in online app stores, where nodes (i.e., reviewers) are connected if they have reviewed the same app.…”

“…Excluding false positives is time-consuming for auditors and can significantly reduce detection efficiency. ML methods mainly use graph neural network (GNN) models to detect collusive fraud [6], [7], [8]. Fraudsters and their associations are constructed as homogeneous or heterogeneous graphs.…”

A Review on Graph Neural Network Methods in Financial Applications

“…Unlike guarantee loans that the loan information could be naturally represented in a directed graph, other loan types may not have a clear graph structure and researchers need to construct the graph based on interactive information. For instance, Xu et al (2021) construct a user relation graph where users are connected by various relationships, such as social connections, transactions, device usage. However, the interactive graph may also contain noisy data, which may be irrelevant.…”

A Comparative Analysis of Bias Amplification in Graph Neural Network Approaches for Recommender Systems

“…Although GNNs usually provide better accuracy in results, most of the existing GNNs do not take the fairness issue into consideration, which can result in discrimination toward certain demographic subgroups with specific values of features that can be considered sensitive, such as age, gender, and race. The decision made by the implemented GNNs can be highly affected by these kinds of discrimination [32,34,35]. In addition, a wide range of ML systems are trained with human-generated data; hence, there is a clear need to comprehend and mitigate bias toward demographic groups in GNN approaches [36].…”