Payment platforms improve fraud detection by combining model results adaptively
InterFuse: Interaction-Guided Adaptive Fusion of Heterogeneous Models for Payment Risk Ranking
Artificial Intelligence
Summary
Payment platforms often use several different models to decide which transactions might be risky and deserve further review. The authors developed InterFuse, a method that smartly combines these models’ predictions by looking at how each model’s information and coverage fit together for each transaction. This approach helps the system decide better which transactions to flag while handling cases where some models might not have full information. Tests show InterFuse improves the accuracy of ranking risky transactions compared to existing combination methods.
What this means in practice
- •For payment platform engineers: Enhance payment risk scoring by adaptively fusing multiple model outputs based on their predictions and data coverage per transaction.
- •For financial fraud analysts: Improve prioritization of suspicious transactions for manual review by combining heterogeneous detection models more effectively.
Authors
Haoyu Jiang, Weiyu Han, Ze Zhao, Jie Sheng, Ruobing Zuo, Xiao Tan, Shuai Chen, Tianyi Zhang
Abstract
Payment platforms use view-specific models trained on transaction attributes, behavioral histories, and relational structures to rank transactions under limited review budgets. However, existing fusion methods often focus on prediction scores or interactions among model representations, with less attention to how these interactions should guide aggregation using each model's original prediction and input coverage. To address this limitation, we propose InterFuse, a fusion framework that uses a shared fusion state from cross-model interaction to guide transaction-adaptive aggregation. InterFuse encodes each model's representation, prediction score, and input coverage as a model-specific token, using learned states for unavailable outputs. Availability-aware attention updates these tokens and produces a shared fusion state. This state then guides aggregation over available models using their updated representations, original predictions, and input coverage. We evaluate InterFuse on an industrial payment replay and three public financial benchmarks using the same upstream models across fusion methods. InterFuse achieves the highest top-ranked recall and AUC on all four datasets, improving Recall@1% by 1.11-8.09 percentage points over the strongest baselines. Ablations show the benefits of cross-model interaction and adaptive aggregation, while industrial analyses find larger interaction gains when model predictions disagree. Additional analyses examine robustness to missing outputs and the trade-off between ranking gains and fusion latency.