A Deep Learning Ensemble With Data Resampling For Credit Card Fraud Detection

Abstract

Credit cards are extremely important in today's digital economy. Its use has been significantly increased in recent years due to credit card theft. Credit card fraud detection has ML classifiers to do their best, using the fact that credit card holders are constantly changing and classroom imbalances are problematic. To solve this problem, this study is a robust methodology of DL with "long short-term memory (LSTM) and Repeat Units (GRU)" as a fundamental student of emergency architecture, and is "multilayer" than meta ears. The technique of hybrid synthetic minority and the modified method of the closest neighbor (SMOTE-EN) is used to increase the distribution of the class in the data file more and at the same time. Experimental findings showed that integration of the proposed DL set with Small-en-Accessible showed that the integration of “sensitivity of 1,000 and specificity of 0.997 recorded other frequently used classifiers and ML methods in the literature. It also covers more complex file models such as stacking and voting classifiers. Test them with both the original and the Small-en data”. The SQLite Integration Framework also allows users to register, register, and test. This will help your project work better and interact with users.