DEFEATFRAUD: Assessment and validation of deep feature engineering and learning solutions for fraud detection

Home > DEFEATFRAUD: Assessment and validation of deep feature engineering and learning solutions for fraud detection

Partners:

 

Funding: TEAMUP program, INNOVIRIS

Duration: April 2018 – March 2020

Project Overview:The project aims at improving the existing fraud detection process of Worldline by adding a number of deep learning and adaptive functionalities to the existing data driven strategies. This will be made possible by increasing the degree of autonomy and adaptivity of the detection process thanks to a number of methodological improvements: 1) design and assessment of an online learning classifier based on deep learning, whose great potential has not yet been explored in the domain of fraud detection 2) automation of the feature creation step by adopting recent representation learning techniques (deep learning) 3) integration of supervised and unsupervised techniques for precision improvement 4) introduction of an exploration step (based on active and semi-supervised learning) in the labeling process to improve the reactivity to fraud change and nonstationarity. We expect several benefits ranging from enhanced fraud detection accuracy, better interpretability of fraudulent patterns, and fraud prevention.