Date of Publication :17th January 2024

Abstract: Electronic payment fraud is a significant concern in today's digital world. Detecting fraudulent transactions accurately and efficiently safeguards financial systems and protects users from financial losses. Due to which we have used Electronic payment fraud is a significant concern in today's digital world. Detecting fraudulent transactions accurately and efficiently safeguards financial systems and protects users from financial losses. We utilized a dataset specifically curated for fraud detection comprised features such as transaction type, amount, balance information, and flags indicating fraud. We performed exploratory data analysis to gain insights into the data distribution and understand the characteristics of fraudulent transactions. Visualizations, including count plots and distribution plots, helped us identify patterns and variations in different features. We employed several algorithms for fraud detection, including Logistic Regression, Support Vector Machines (SVM), XGBoost, and Naive Bayes, The analysis revealed varied model performances. Logistic Regression and SVM achieved 100% accuracy. XGBoost showed higher accuracy at 100%, while Naive Bayes achieved 41%. Random Forest outperformed others with 100% accuracy with minimum losss. These findings highlight the variability in performance, with Random Forest emerging as the most effective model. Logistic Regression, SVM, and XGBoost also demonstrated excellent accuracy levels.

Reference :