Bootstrapping the Paysim Financial Simulator for Open Source

Abstract

PaySim Simulator is an approach to the the lack of legitimate datasets on mobile money transactions to perform and experiment with fraud detection techniques. In this paper we used a technique called bootstrapping which uses the synthetic dataset generated by PaySim to build the parameter files that were initially extracted as aggregated information from the original data which is sensitive to share to the public. This way, researchers will have the possibility to generate diverse dataset with sufficient quality to perform fraud detection research on it. This paper explains the bootstrapping approach and in addition a method to properly reuse the PaySim simulator in the same or similar other domains. This is a solution to ultimately yield the possibility to simulate and generate financial transactions in such a way that they become similar to the original dataset without any direct connection to the original source.