Improvements in Fraud Prevention using Machine Learning

Credit card fraud is becoming a concern for both individuals and financial organizations due to the rapid expansion of online financial transactions. Our research presents one technique for spotting credit card fraud. Our suggested solution uses a combination of algorithms and under monitoring to quickly and accurately identify current fraudulent activity. We assess the system's performance using important metrics, such as F1 score and accuracy on a dataset containing both valid and illicit transactions, to make sure we examine the system's efficacy. The methods used to determine which credit cards are fraudulent includes logistic regression, kNN, Random Forest, Support Vector Machine (SVM), MLP model, SGD model, and Extra Tree model. The startling finding is that more than 70% of respondents acknowledged having one or more credit cards. Regretfully, as credit cards have become more popular, so has their use. There are two types of these fraudulent acts. First, an identity thief registers an ATM card account in the victim's name; figures show that between 2019 and 2020, this type of fraud increased by a startling 48%. The second kind of deception, which occurs when an identity thief obtains access to a person's current credit card account, was reported 9% more frequently in 2019 and 2020. The concerning increase in credit card fraud emphasizes the need for further security measures to safeguard consumers’ financial security. The analysis of the exploratory work's findings reveals a notable improvement in the ROC curves with optimized values, delicacy, perfection, and F1-score. This prompted research into a number of machine learning approaches and tactics to address the issue of identifying credit card fraud.