Machine Learning For Detecting Credit Card Fraud

Abstract

These days frauds related to credit cards are exponentially increasing as compared to earlier scenarios. Like every coin has two faces in a similar way where on one hand the introduction of credit cards has helped in the ease of online payment to make our lives easier, on the other hand, the same technology has increased the number of frauds. Fake identities and various technologies are used by the criminals or cyber attackers to trap the users. Henceforth, it has become essential to find a solution for all such frauds and abnormal activities so that money of the user can be protected at the time of transaction. In order to tackle all such problems, we can train our machines using Machine Learning Algorithms. This project has been designed to illustrate the analysis of the dataset taken from Kaggle and train our system accordingly so that any kind of abnormal activity during a transaction can be immediately detected. The issue involves examining previous credit card transactions using information from both the fraudulent ones and the zeroes that were legitimate Here, detecting 100% of fraudulent transactions while lowering erroneous fraud classifications is our key goal. Data sets are analyzed and pre-processed, and various anomaly detection techniques, like the Random Forest algorithm and Decision Tree Classifier, are used to get the Prompt Corrective Action modified Credit Card Transaction data, have been the main focuses of this procedure. The models are compared and evaluated based on training and testing accuracy. It has been found that the Decision tree classifier performed better on training accuracy i.e., 95% while random forest demonstrated better testing accuracy i.e., 94.11%.