A Customer-Level Fraudulent Activity Detection Benchmark for Enhancing Machine Learning Model Research and Evaluation

Aug 4, 2026

Upload by Noah

Fraud detection research depends on comprehensive, privacy-compliant data to train and evaluate machine learning methods that can identify sophisticated schemes. Existing resources mainly capture transaction-level records while neglecting customer behavior patterns, and privacy restrictions further limit dataset availability. This study introduces a structured customer-level benchmark dataset containing customer-centric features over time and supporting model assessment. The benchmark enables evaluation of diverse ML approaches, including deep learning and anomaly detection, in realistic scenarios, advancing safer anti-fraud systems. Code and data are planned to be open-source.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Benchmark

A Customer-Level Fraudulent Activity Detection Benchmark for Enhancing Machine Learning Model Research and Evaluation Phoebe Jing, Yijing Gao, Xianlong Zeng1 Abstract In the field of fraud detection, the availability of comprehensive and privacy-compliant datasets is crucial for advancing machine learning research and developing effective anti-fraud systems. Traditional datasets often focus on transaction-level information, which, while useful, overlooks the broader context of customer behavior patterns that are essential for detecting sophisticated fraud schemes. The scarcity of such data, primarily due to privacy concerns, significantly hampers the development and testing of predictive models that can operate effectively at the customer level. Addressing this gap, our study introduces a benchmark that contains structured datasets specifically designed for customer-level fraud detection. The benchmark not only adheres to strict privacy guidelines to ensure user confidentiality but also provides a rich source of information by encapsulating customer-centric features. We have developed the benchmark that allows for the comprehensive evaluation of various machine learning models, facilitating a deeper understanding of their strengths and weaknesses in predicting fraudulent activities. Our dataset includes detailed attributes that reflect customer behavior over time, thereby enabling researchers to employ and test advanced analytical techniques, such as deep learning and anomaly detection algorithms, under realistic scenarios. By offering the dataset to the research community, we aim to set a new standard in fraud detection research, providing a tool that can significantly enhance the predictive accuracy of fraud detection systems. This initiative not only fosters innovation in machine learning model development but also contributes to safer banking practices, ultimately protecting consumers and financial institutions alike from the perils of fraudulent activities. Through this work, we seek to bridge the existing gap in data availability, offering researchers and practitioners a valuable resource that empowers the development of next-generation fraud detection techniques. Our code and data will be open-source and publicly available at Github. Introduction Fraud detection remains a critical challenge in the banking sector, where the ability to quickly and accurately identify fraudulent activities can significantly influence the financial stability of [1](1xz926813@ohio.edu)[xz926813@ohio.edu](1xz926813@ohio.edu) institutions and the security of customer assets. Machine learning ( ML) has emerged as a pivotal tool in combating fraud, leveraging vast amounts of data to discern patterns and anomalies that human analysts might miss. The efficiency of ML in fraud detection is largely contingent upon the quality and scope of the data used. However, while these systems are potent in theory, their practical efficacy is often hindered by the limitations inherent in the available datasets. Most existing datasets for fraud detection are primarily focused on transaction-level data. These datasets typically provide detailed information about individual transactions but lack broader behavioral contexts that might provide insights into customer-level patterns of fraudulent behavior. This focus can lead to significant gaps in understanding and predicting fraud because it fails to capture the cumulative anomalies or consistent patterns of behavior at the customer level, which are often more indicative of fraud. Furthermore, the availability of these datasets is significantly constrained by privacy concerns. Banks and financial institutions must adhere to strict data protection regulations, such as the GDPR in Europe and various other privacy frameworks globally, which limit the sharing and utilization of sensitive customer data for research purposes. To address these challenges, our study proposes the development of a structured

FAQs

Why do traditional fraud detection datasets limit machine learning performance?

They mainly focus on transaction-level information and omit broader customer behavior context that often reveals consistent or cumulative fraud patterns.

What does the proposed benchmark provide to researchers?

It provides structured customer-level datasets with privacy compliance and customer-centric features capturing behavior over time.

How can the benchmark be used to evaluate machine learning models?

Researchers can run models against the benchmark to compare performance using metrics relevant to practical fraud detection, such as accuracy and recall/precision.

Aug 4, 2026

Upload by Noah