Identification of Fraudulent Transactions for Enhanced Credit Card Fraud Detection

With the flourishment of the digital era and the evolution of new techniques and technology, this paper highlights the ever-growing risk of credit card fraud. The advancing technology has provided fraudsters with new and untraceable techniques for scamming genuine users. With the increasing impact of fraud, financial institutions must use machine learning tools for scam detection to provide safety and security to their users' accounts. This paper consists of an interdisciplinary comparative analysis of ML algorithms: Random Forest Classifier, XGBoost, ADABoost, Logistic Regression, Decision Trees, and Artificial Neural Networks. The diverse selection of algorithms aims to determine the most effective algorithm with the highest accuracy score for predicting legitimate and fraudulent transactions. The methodology involves fetching data from an Excel file, training the models, testing them on 20% of the dataset, and finally obtaining results in a Word document, having a confusion matrix and accuracy of each algorithm. By comparing the results of the confusion matrix and accuracy score, the Random Forest Classifier and XGBoost demonstrate the highest accuracy scores, while Logistic Regression yields a lower accuracy score.