Table 3 Justification of architectural design.
From: RABEM: risk-adaptive Bayesian ensemble model for fraud detection
Component | Role in RABEM-Fraud |
|---|---|
Black-scholes features | Adds domain-aware financial risk metrics (e.g., volatility) to enhance feature space |
VAE | Performs unsupervised anomaly detection, capturing rare fraud behavior patterns |
Nyström GPC | Adds probabilistic modeling with uncertainty quantification; scalable via approximation |
RPTree | Lightweight high-dimensional classifier with fast inference |
GRU | Captures temporal fraud sequences and evolving transaction behavior |
Bayesian fusion | Dynamically combines predictions using model reliability and entropy |
