Enhanced GAN based Fair Synthetic Data Generation Model

Abstract

AI based models are increasingly touching the human life. Critical decisions such as credit limit allocation, loan approval etc. are being increasingly taken with the help of advanced AI/ ML models. The accuracy of these models has increased tremendously over the years. However, newlinethese models are also data hungry. For instance, neural network-based models normally require huge amount of data for fitment/ training and validation. Availability of data has often remained newlinea challenge with these AI/ ML based models. Further, various regulatory requirements and newlineethical considerations are making it increasingly important for AI/ ML based models to produce fair output as well along with high accuracy. Hence, an effective synthetic tabular data generation technique needs to meet the fairness objective as well apart from being similar to the original data. newline