Performance Comparison of Credit Card Fraud Detection System using Machine Learning
Alavikunhu Panthakkan, Najiya Koderi Valappil, Majida Appathil, Seema Verma, Wathiq Mansoor, Hussain Al-Ahmad · 2022
Cyber security is becoming an integral part of modern life, tackling abnormal activities being the main challenge in the domain. To find and abort a fishy transaction is the procedure for maintain the security in credit cards. With the availability of past datasets, machine learning algorithms have made it possible to detect abnormal activities among the transactions. The aim of the paper is to create a balanced data from a pre-existing dataset and to apply six different machine learning algorithms like Decision tree, K-Nearest Neighbor, Logistic regression, SVM, Random Forest and XGBoost to detect fraud activities. Later, the results are compared and discussed on the basis of evaluation matrices which include precision, recall, f1-score and accuracy. All the algorithms are applied on three sets of data, where the ratio of fraud to valid transactions is varied. Three cases are considered with ratios as 1:100, 1:10 and 1:1. A balanced dataset showed greater performance in the terms of all evaluation matrices.