Feature Construction by Using Autoencoder’s Fusion for Credit Card Fraud Detection

Authors

  • Deepika.S, S.Senthil

Abstract

From past few years fraud transactions in credit card transaction is drastically increased in banking and e-commerce organizations. Hence need an automatic fraud detection model which can address all issues of customer. Machine Learning is one of the solutions to design model to find credit card fraud detection through the use supervised binary classification approach. In machine learning, the accuracy is depends on feature sets. In this paper we derived more number of feature sets from original feature set using different normalizations strategies. By using auto-encoders fusion, derived most optimal feature set and then used in Generative Adversarial Network (GAN) for training and to identify fraud transaction. These normalized optimal feature vectors with auto encoder fusion and GAN model given good results when compared with other state-of-art methods.

Downloads

Published

2020-05-12

Issue

Section

Articles