SINDy-Augmented Hybrid Fraud Detection Artifacts (PaySim)

This archive contains code, processed datasets, trained models, configuration files, and evaluation metrics accompanying the paper “Hybrid Fraud Detection with Sparse Identification of Nonlinear Dynamics–Augmented Dynamic Features LightGBM-Based Performance and Explainability”.The original PaySim dataset is not redistributed; instead, we provide instructions to download it and share only derived Parquet tables, dynamic SINDy models, and classifier artifacts. See the README in the archive for a detailed mapping to the figures and tables in the manuscript.