---
title: Comparative Evaluation of Anomaly Detection Methods for Fraud Detection in Online Credit Card Payments
url: https://www.ml-quant.com/papers/arxiv/2312.13896/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-09-26
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
identifier: arXiv:2312.13896
source_url: http://arxiv.org/abs/2312.13896
featured: 2024-01-03
citations: 16
topic: Risk, Credit & Banking
---


# Comparative Evaluation of Anomaly Detection Methods for Fraud Detection in Online Credit Card Payments

A study found that LightGBM was the best for fraud detection when comparing anomaly detection and standard supervised learning methods, but it was more susceptible to distribution shifts, questioning the advantage of combining these two methods.

- Source: http://arxiv.org/abs/2312.13896
- Identifier: arXiv:2312.13896
- Released: 2023-12-21
- First featured: Quant Letter No. 31 (2024-01-03): https://www.ml-quant.com/issues/2024-01-03/
- Citations (Semantic Scholar): 16
- Published in: not yet
- Topic: Risk, Credit & Banking

## Related

- [Machine and Deep Learning for Credit Scoring: A compliant approach](https://www.ml-quant.com/papers/arxiv/2412.20225/): The research proposes new BASEL 2 and 3 compliant techniques for credit scoring in banks, demonstrating improved performance and default capture rate with Gradient Boosting Machines.
- [Re-examining Standard Changes in U.S. Corporate Credit Rating via Machine Learning](https://www.ml-quant.com/papers/ssrn/4771505/): Machine learning reveals an N-shaped pattern in corporate credit rating standards from 1986 to 2016, with the Gradient Boosting Machine model predicting actual credit ratings better than traditional regression models.
- [Scaling SMEs Credit Scoring](https://www.ml-quant.com/papers/repec/hal-wpaper-hal-04159788/): A new method using Gradient Boosting Decision Trees and SHapley Additive exPlanation values aims to enhance credit scoring for Small and Medium Size Enterprises, providing high predictability and explainability.
- [Comparative Credit Risk Model Analysis](https://www.ml-quant.com/papers/ssrn/5204562/): The analysis reveals that gradient boosting models, specifically CatBoost and LightGBM, are more effective than traditional models in assessing credit risk.
- [Defining and comparing SICR-events for classifying impaired loans under IFRS 9](https://www.ml-quant.com/papers/arxiv/2303.03080/): The paper presents a new framework for predicting credit deterioration using three parameters, validated using South African mortgage data.
- [An Explicit Scheme for Pathwise XVA Computations](https://www.ml-quant.com/papers/arxiv/2401.13314/): A new simulation/regression scheme for a type of anticipated BSDEs is introduced, using neural network least-squares and quantile regressions, showing better results in high-dimensional and hybrid market/default risks XVA use-case.