cff-version: 1.2.0 message: "If you use this software or its findings, please cite it as below." title: "fraud-synth-bench: Reproducible fraud-analytics research on synthetic data" abstract: >- A seeded synthetic transaction generator with pluggable fraud scenarios, drift schedules and label delay, plus a pre-registered comparison of classic and modern fraud-detection models under extreme imbalance, concept drift, ring fraud, label delay and investigator capacity limits. type: software authors: - family-names: Chongmankhong given-names: Thanatchon license: Apache-2.0 repository-code: "https://github.com/thanatchon36/fraud-synth-bench" doi: 10.5281/zenodo.23343483 version: 0.1.1 date-released: "2026-10-13" keywords: - fraud detection - synthetic data - benchmark - reproducibility - gradient boosting - graph neural networks