Published Paper Details:

Credit card theft has increased as e-commerce has expanded. Banks are having a harder time identifying credit card fraud as a result of the industry's explosive growth. One of the most crucial tasks that machine learning performs in preventing credit card fraud is the identification of fraudulent purchases. Banks estimate these types of transactions using a range of machine learning techniques, building on historical data and adding special characteristics to boost prediction accuracy. The strategy for sampling the data set, the choice of variables, and the detection process are all crucial to the effectiveness of credit card fraud detection. The usefulness of decision trees, random forests, and logistic regression in identifying credit card fraud is examined in this study. A dataset of 2, 84,808 credit card transactions from a European bank collected using kaggle. It defines fraudulent transactions as "positive class" and valid purchases as "negative class," but the data set is considerably skewed, with just about 0.172% being fraudulent and the rest being legitimate because to the author's frame interpolation efforts, the dataset now contains 60% fraudulent and 40% valid transactions. The dataset is subjected to each of the three approaches, and the output code is written in the programming R. Several metrics, including as sensitivity, specificity, accuracy, and error rate, are utilized to analyze the efficacy of the approaches in relation to the aforementioned criteria. The accuracy of the logistic regression, decision tree, and random forest classifiers was 90.0, 94.3, and 95.5%, respectively. When evaluated to logistic regression and decision trees, the Random forest outperforms both.