Generative AI in Financial Fraud Detection

Abstract

the objective of this exploration is to utilize AI calculations to recognize Visa misrepresentation. Because machine learning algorithms can examine enormous datasets and uncover patterns suggestive of fraud, they present a promising method for detecting fraudulent transactions. An experimenta design is used in the research approach to gather, prepare, and analyze data. The study's dataset, which included 23 variables linked to financial transactions, was sourced from Kaggle. Feature engineering, addressing missing values, and removing superfluous columns were all part of the data pre-processing step. Three machine learning models — the Random Forest Classifier, K-Nearest Neighbor (KNN), and Support Vector Machine (SVM) — were prepared and evaluated. The models were evaluated utilizing performance measures such area under the curve (AUC), review, exactness, accuracy, and F1-score. The review's discoveries show that Random Forest and KNN beat SVM in the distinguishing proof of Mastercard fraud. With an exactness of 99.63%, accuracy of 99.53%, review of 99.45%, and F1-score of 99.80%, Random Forest performed very well. With an exactness of 99.45%, accuracy of 99.87%, review of 99.89%, and F1-score of 99.80%, KNN performed all around well. With a F1-score of 99.61%, an exactness of 99.58%, accuracy of 99.74%, and review of 100 percent, SVM performed all around well. As per the examination, machine learning calculations might be helpful in recognizing Mastercard fraud. The capacity of machine learning calculations to recognize Visa theft is shown by this review. The results exhibit that Random Forest and KNN perform preferable on this task over SVM. The outcomes offer critical viewpoints for financial foundations and endeavors seeking to further develop their fraud detection systems. To expand the exactness of fraud detection, future examination could focus on exploring elective machine learning calculations and systems.