An Algorithmic Approach to Machine Learning Techniques for Fraud detection: A Comparative Analysis

Abstract

Fraud using credit cards is still rife today, and the modes are increasingly varied. To avoid scams with various ways of credit cards, we must identify and find out what methods are often used by fraudsters. The comparative analysis depicts that the parameters, i.e., Precision/Recall and F1-Score the K-Nearest Neighbor, are better for detecting fraudulent transactions than the Logistic Regression and Naïve Bayes. However, the accuracy is marginal high of Logistic Regression, but the False Positive parameters cannot identify the imbalanced data; therefore, they disguise the results and accuracy of Logistic Regression and K--Nearest Neighbor deems fit for such cases. Kaggle Dataset for fraud detection has been used to experiment. Therefore, under the scheme, we used various models of machine learning models based on classification and Regression. The results show that the K--Nearest Neighbor is the better approach for detecting fraudulent transactions than the Logistic Regression and Naïve Bayes.