AI driven fraud detection in digital payments using big data and machine learning
DOI:
https://doi.org/10.5281/zenodo.15968517Keywords:
Fraud Detection,Machine Learning,Digital Payments,Anomaly Detection,Big Data Analytics,Artificial Intelligence,Transaction Monitoring,Behavioral Analytics,Real-Time Detection,Predictive Modeling,Risk Scoring,Neural Networks,Supervised Learning,Unsupervised Learning,Data Stream Processing.Abstract
This paper reviews the investigation into the fraudulent actions undertaken with the help of big data. Notably, the analysis of credit compatibility and classification has grown increasingly intriguing in recent years due to the rapid increase in value and data all around the world. Nowadays, the electronic channel is ruled by industrial changes owing to connected card processing and online purchasing. This innovation has plenty of usage as a result of its accessibility and get hold of the world. As the number of consumers rises daily, controlling credit card fraud becomes a looming danger to banks and financial agencies because it costs a ton of money. Credit card fraud tends to happen during both online and offline purchasing; the conditions are barely different, and the consequences are immense nowadays. For any credit card fraud to be successful, the cardholder's personal details are needed, especially the 16-digit card number consisting of the card provider's identification number, card account number, and card industry identification number. Since card datasets are extremely unbalanced. Also, a picture of the card and a CVV pin may help. The goal of this undertaking is to reduce the number of false warnings at merchant and bank payments and the number of false denials. False warnings have already been published every year, resulting in the merchants losing around $118 billion and the customers losing around $9 billion. This is fraud in finance.
Machine learning is a trump card as a whole. Algorithms can interpret a semi-pre-arranged dataset along with different features. Such algorithms are fast, maintainable, and less human-dependent and can handle various types of data. As a result, replicating the behaviour of a person, such as credit card usage or approval, cellphone direction, or other perceptions, is simpler. These models can be applied in web services, and apps and hardware have already fled into the market, catching bank thieves. Several gathering models are involved in consuming data through appliance learning, particularly when it is fragmented into working sets, such as cluster analysis and anomaly detection. A major worry to be dispersed during banking fraud is the more advanced and diverse varieties of banks due to online data fragmentation.
