Harmonized Multi-Source Fraud Transaction Dataset for Benchmarking Synthetic Data Generation Methods

Description

This repository contains a harmonized multi-source fraud transaction dataset integrating IEEE-CIS, ULB, and PaySim datasets. The dataset contains 5,000 balanced transactions (2,500 legitimate and 2,500 fraudulent) with standardized time, amount, transaction type, fraud label, and source information. It also includes source datasets, temporal validation splits, and the harmonization notebook to support reproducible fraud-detection research and benchmarking.

Steps to reproduce

Download the harmonized dataset and supporting files from this repository. Refer to the data dictionary and feature harmonization specification to understand the variables. Use the harmonized dataset (harmonized_fraud_dataset_5000.csv) for fraud detection experiments. Use the provided SMOTE, CTGAN, and WGAN datasets to benchmark synthetic data generation methods. Compare model performance using standard evaluation metrics such as Accuracy, Precision, Recall, F1-score, ROC-AUC, and PR-AUC as described in the accompanying Data in Brief article.

Institutions

• Symbiosis International University

  Maharashtra, Pune

Categories

Licence