International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 10 Issue: 05 | May 2023
p-ISSN: 2395-0072
www.irjet.net
Credit Card Fraud Detection Using Machine Learning Zainab Firdous1, Sushma V2, Aftab Pasha S3, M Shahista Banu4, Najmusher H5 1,2,3,4Dept. of CSE, HKBK College of Engineering, Bangalore
5Professor, Dept. of CSE, HKBK College of Engineering Bangalore
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – Fraud is the act of depriving a
performance of various algorithms such as logistic regression, decision tree, random forest and support vector machine based on precision, recall and accuracy. The traditional systems available such as CIBIL score uses demographic data, credit history and so on to calculate a score. This score is employed by many money lending organizations to judge whether to issue the loan to this applicant or not and to set the credit limit in case of credit cards. However, this system only gives an idea to banks of the level of risk involved in granting the person loan. Hence, we have developed a model that predict with precision the probability of a user defaulting.
person/organization of money through willingness, deception or other unfair means. The unforeseen event of Covid-19 has led to many people embracing digital transactions and online shopping. This combined with other benefits provided by credit card issuers such as rewards has increased the usage of credit card, and in turn increased credit card frauds. Credit card default can have serious implications on credit card holder and can affect financial stability of credit card issuers. There is a need for a system that can predict defaults ahead of time so that appropriate measures can be taken by credit card issuers. In this paper, we have provided a comparative analysis of various machine-learning algorithms often used in fraud detection such as logistic regression, decision tree classifier, random forest classifier and support vector machine classifier. The models were compared on the basis of precision, recall and accuracy to find the best algorithm for predicting probable defaulters.
2. LITERATURE REVIEW The paper by author Yue Yu, [1] concentrates on the importance of credit card default prediction by highlighting the need of timely and precise prediction to prevent financial losses for both banks and the users, then proceeding to make an outline of various machine-learning based algorithms by explaining their principles, advantages and comparing their performances. Yue Yu then introduces the dataset used in the paper, which includes thirty thousand entries of credit card users from a Taiwanese bank containing their previous credit card transaction information, card-user demographics, and payment behavior to train and evaluate these algorithms. The paper then presents an analysis of applying the selected machine learning algorithms to the dataset and evaluating their performance using parameters such as accuracy, precision, recall, score of harmonic mean and Receiver Operating Characteristic graph. The results show that all the algorithms considered for the test produce good results but artificial neural networks and support vector machine giving better accuracy and performance rates. The article ends by admitting some drawbacks and potential scope for future study.
Key Words: Credit Card fraud, Credit Card default, Machine Learning, Support Vector Machine, Decision Tree, Logistic Regression, Random Forest
1.INTRODUCTION The banks earn money by various means such as lending loans to other customers using the depositor’s money. The interest gained is the profit earned by the bank, but when the borrower defaults, the loan becomes a NPA and is a huge blow to the bank’s statements. It was estimated that the total amount of NPAs in India increased from 2.39 lakh crore in 2014 to 10.36 lakh crore in 2018. Therefore, an effective system must be developed to curb these defaults even before they occur. Different types of defaults in finance include:
Loan default: This is the most common type of default. In this case the loan borrower fails to repay the loan. Credit card default: This occurs when the credit card holder uses his/her credit card to buy items that they cannot afford but doesn’t repay the money spent. Bond default: when the organization/government fails to repay the loan/principal amount, it is considered as bond default.
Authors, Yashna Sayjadah, Khairl Azhar Kasmiran, Ibrahim Abaker Targio Hashem, and Faiz Alotaibi, [2] in their paper, have recognized challenges associated with credit card default prediction, such as imbalanced datasets and the requirement for precise models. To analyze their study, the authors obtained a dataset containing credit card information and default status which is pre-processed by handling missing values, encoding certain variables, and conducting feature ascending. Following which the dataset is categorized into training and testing purposes. Accuracy of different algorithms was noted for the possible occurrence of default credit card. Algorithms used here are logistic
We have focused on developing a system to predict credit card defaulters. A comparative analysis was done on the
© 2023, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 1478