Enhanced GANs Approaches for Unbalanced Data in Risk Management

Abstract

Fraudulent transactions as well as fund stealing create enormous difficulties for banking organizations as well as society as a whole. Financial institutions devote significant efforts to detecting and combating suspicious and illegal activity. Large institutions are said to have saved 150 million dollars in just one fiscal year by using AI to identify scams. Existing methods for detecting fraud depend on human-engineered rule databases which correlate with odd patterns in monetary transactions. Additional regulations are introduced to the rules database when novel schemes are discovered. The strategy is to use GANs to simulate these difficulties for unbalanced datasets in anomaly detection. Imbalanced data categorization is a difficult problem for Data Scientists. The researchers in this work assume that the dataset contains 99.9% normal transactions and 0.1% fraudulent transactions.