January-February 2020

ISSN: 0193-4120 Page No. 10324 - 10327

Credit Card Fraud Detection System Using

Machine Learning Process
*1M. Sai Vikas, 2S. Rinesh, 3D. Mahalakshmi
*1UG Scholar, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences,
Chennai
2,3
Assistant Professor, Saveetha School of Engineering, Saveetha Institute of Medical and Technical
Sciences, Chennai
*[email protected], [email protected], [email protected]

Article Info Abstract

Volume 82 In recent years, machine learning has been widely used for the fraud detection
Page Number: 10324 - 10327 process and achieved favorable performance. According to the Financial sectors
Publication Issue: have focused attention to recent computational methodologies to provide the
January-February 2020 credit card fraud problem. Our analysis provides a comprehensive guide to
sensitivity analysis of current parameters with regards to words the current
performance in credit card fraud detection. It defines only the numerical input
variables which the help of the Principal Component Analysis (PCA)
transformation. Unfortunately, due to confidentiality issues, we should not
provide the original features and more background information to be provided.
To predict machine learning model to predict whether a transaction is fraudulent
or not by approaching Logistic, Support vector classifier, Random forest
algorithms and identify the most important variables that may lead to higher
accuracy in credit card fraudulent transaction detection. Additionally, we can
compare and discuss the performance of various machine learning algorithms
from the bank credit dataset with evaluation classification report from Principal
component analysis and identify the confusion matrix and scalar metrics. So,
Article History present a framework of the parameter of the Machine learning topologies for the
Article Received: 18 May 2019 credit card fraud detection is to be enable financial institutions to reduce losses
Revised: 14 July 2019 by preventing fraudulent activity to words the bank related process.
Accepted: 22 December 2019
Publication: 19 February 2020 Key words: Keywords: dataset, Machine learning- Supervised classification
method, Accuracy prediction

1. Introduction someone else‟s credit card for their personal

The fraud in credit card transactions is reasons while the card holder and the card
inaccessible and unauthorized usage of an issuing Agencies are unknown of the fact that
account by others than the owner of that current the card is being used by someone else.
account. Necessary steps could be taken to stop The fraud detection mainly involves
this type of behavior and the fraudulent monitoring of database and the activities of
practices can be studied to reduce it and to their customers, in order to keep away fraud
protect against this type of occurrences in the behaviors, default frauds and interruptions.
future. In further cases, the Credit Card Fraud This is a relative problem that demands the
can be defined as a process of misusing Communities attention such as machine

Published by: The Mattingley Publishing Co., Inc. 10324

 January-February 2020
ISSN: 0193-4120 Page No. 10324 - 10327

learning and Artificial intelligence where we should be taken accurately. Previously at

can find the solution to the problem. This type particular intervals we have used effective
of problem is especially challenging from measures to increase the fraud detections. The
learning aspect, as it can be described through main cause for this Credit Card frauds is lack of
various factors such as class asymmetry. There security and lack of knowledge about banking
is a chance that the transaction patterns even to the user. Even though it is not an effective
change the statistic property in time and the way of fraud detection it is quite simple way
number of transactions processed should be with high ranging performance to find the
invalid. details of the fraud. Now a days we are using
Machine learning mechanisms are equipped various which are of low security the virtual
to access the authorized transactions and find proposed here is to overcome from these
out the fraud detections. The transaction details processes.
will be controlled by the authorities and their 2) Online Credit Card Detection Using Machine
representatives contact the account holders to Learning To Rank Approaches [2]
confirm whether the transaction is real or fraud. Furtherly by using internet dealing, the payment
is formed wherever the cardboard holder doesn't
2. Literature Review need to physically manufacturing the card to the
1) GUI Based Credit Card Fraud Detection business person for confirming its originality at
processes [1] the time that associate order is made and
In day to day life Credit Card frauds are payment is underneath method, like for getting
becoming a major problem and our aim is to associate item through an e-commerce web site
overcome this problem before the fraud is via web banking. It‟s terribly crucial to verify
happened. There are various methods to find out that associate authentic cardholder is so activity
these frauds Neutral networking, a dealing just in case of a card not gift state of
HiddenMarkoff model. This technique involves affairs to form positive that the transaction is
Net banking login and a onetime registration non-fraudulent. Most of the businesses and
process. Once the details are being processed industries have remodeled their business into
with Hidden MarkoffModel, the Fraud on-line services to supply ecommerce, easy
Detection system controls the Card holders accessibility and communication permitting
account details and there by taking fresh style their customers higher potency and
formula into consideration we are able to find accessibility. This evolution is of nice
the average. Previously we mainly targeted on importance since it's attributable towards
detecting fraud transactions. But using this enhanced potency, and gain however it
formula we are able to find out the details of the additionally has its own shortcomings. Along
fraud happened with accurate details. In initial with this evolution comes a bigger chance of
stage we tried to overcome the fraud by threats. One in every of the foremost crucial
providing necessary credentials about the challenges in creating business on the web is
account to the authority and further they will that neither the cardboard nor the cardholder
proceed which is a time taking process. The must be gift once the acquisition is being
details of each case will be tested by using created. This makes it not possible for the
Hidden Markoff Model and Genetic formula. business person to verify whether or not the
By the help of these strategies the averages client creating an acquisition is that the
authentic cardholder or not. victimization the

Published by: The Mattingley Publishing Co., Inc. 10325

 January-February 2020
ISSN: 0193-4120 Page No. 10324 - 10327

MasterCard fraud detection engine that, have  A time limit can be set for the applicant to
planned, MasterCard frauds will be prevented check whether his/her credit transaction can be
by lease the businessperson attentive to the fraud or not and to provide quick, immediate
transactions having a high chance of being and easy way to choose the deserving
deceitful. Victimization data processing applicants.
technique alongside neural network, the success To reduce the risk factor by selecting the safe
rate of the system will increase several folds person so that we can save lots of bank efforts
and therefore can solve the aim of aiding the and assets by this process.
merchants. Though there are existing systems
doing a similar factor presently within the 4. Result and Discussion
market, this technique being developed will Table 1: Performance analysis for seven
significantly target the businessperson facet of different algorithms
the business which is able to be helpful to the Detection False
merchant by reducing the merchant‟s losses that Rate Alarm
he should bear if dealing is deceitful. Techniques Accuracy (Precision) Rate
Support
Vector
3. Proposed System Machine
Basic methods to detect the frauds in the credit (SVM) 95.65% 86.45% 6.20%
card system. Comparisons are made with Artificial
Neural
various machine learning algorithms should
Networks
used in it, such as Logistic regression, Decision (ANN) 98.71% 98.68% 0.22%
trees, and Random forest should be done in it. Bayesian
Network 96.52% 96.04% 1.50%
K- Nearest
Neighbour
(KNN) 96.15% 95.84% 1.88%
Fuzzy
Logic
Based
System 94.20% 85.84% 1.25%
Decision
Trees 96.93% 97.52% 3.19%
Logistic
Regression 93.70% 76.80% 3.90%

Figure 1: Proposed System

Advantages
 The goal of this problem is to predict the
status of the credit card fraud detection in credit
card transaction of test dataset as accurate as
possible.

Published by: The Mattingley Publishing Co., Inc. 10326

 January-February 2020
ISSN: 0193-4120 Page No. 10324 - 10327

to the random forest classifier is better than the

Logistic regression and decision tree process
should done.

Reference
[1] Credit Card Fraud Detection Based on
Transaction Behaviour -by John Richard D.
Kho, Larry A published by Proc. of the
(2017).
[2] Research on Credit Card Fraud Detection
Model Based on Distance Sum – by Wen-
Figure 2: Average execution time of the algorithm
Fang YU and Na Wang published by (2009)
Result International Joint Conference on Artificial
Intelligence.
Now a day‟s fraud detection techniques are [3] Credit Card Fraud Detection through Par
highly improved in order to reduce we are enclitic Network Analysis By Massimiliano
implemented multiple machine learning Zanin, Miguel Romance, Regino Criado,
algorithm techniques are implemented to reduce and Santiago Moral” published in (2018).
the fraud detection process. Mostly the fraud [4] “Credit Card Fraud Detection: A Realistic
detection process should done in bank related Modeling and a Novel Learning Strategy”
sectors. To reducing the fraud detection process, published by IEEE TRANSACTIONS ON
we have improved some security features are NEURAL NETWORKS AND LEARNING
SYSTEMS in (AUGUST 2018)
implemented in this process. And finally it
[5] “Credit Card Fraud Detection-by Ishu
shows the „0‟ means fraudulent transactions &
Trivedi, Monika, Mrigya, Mridushi”
„1‟ means Non fraudulent transactions. published by International Journal of
Accordingly, we find the transaction process is Advanced Research in Computer and
safe or not by executing the machine learning Communication Engineering in (January
algorithms techniques. 2016).
[6] Kalaiselvi N, Mahalakshmi S and
5. Conclusion Padmavathy J, karthiga Joyce.B Credit Card
Fraud Detection Using Learning To Rank
In this Current paper, Machine Learning
Approach in the year(2018)
technique are used for the Logistic regression
[7] Marc Rippen, Harry Braunstein and Craig
process, Decision Tree and Random forest Nelson Interactive Multitask Credit Card
algorithm should be done in it, In order to Technology in (2007).
identify the Fraud process should be used in the [8] Wei Li, Xuemei Wu and Yayun Sun Credit
Fraud detection process in the Credit card Card Customer Segmentation and Target
system. Sensitivity, Specificity, Accuracy and it Marketing Based on Data Mining in (2018).
shows the error rate should be used to calculate
the performance of the proposed system
process. The accuracy of logistic regression,
Decision tree and random forest classifier are
91.0, 92.3, and 96.5 respectively. By developing
all these three processes should found according

Published by: The Mattingley Publishing Co., Inc. 10327