CREDIT CARD FRAUD DETECTION

USING MACHINE LEARNING

Hasir Ahmad Janani Kayalvizhi

Sri Shanmugha College of Sri Shanmugha College of Sri Shanmugha College of
Engineering and Technology Engineering and Technology Engineering and Technology
Sankari Sankari Sankari
hasirahmad124@gmail.com jananict8@gmail.com kayalkayu2003@gmail.com
ABSTRACT

When the world was under lockdown and movement was restricted to an
absolute emergency, millions were introduced to the world of online
shopping. The convenience of online shopping helped e-commerce
platforms record historic sales. While that happened, it is no surprise that
the rate of online financial fraud also increased incredibly. Online fraud
cases using credit and debit cards saw a historic upsurge of 225 percent
during the COVID-19 pandemic in 2020 as compared to 2019. As per the
NCRB report, the tally of credit and debit card fraud stood at 1194 in 2020
compared to 367 in 2019. As per figures revealed by the Reserve Bank of
India (RBI), in response to an RTI request, on average, 229 bank frauds
were committed every day in India, resulting in transactions worth Rs.
1.38 lakh crore being syphoned off. The recovery rate for the same has not
been impressive either.

1 INTRODUCTION
Credit card fraud is a term that has been coined for unauthorized access of
payment cards like credit cards or debit cards to pay for using services or
goods. Hackers or fraudsters may obtain the confidential details of the
card from unsecured websites. When a fraudster compromises an
individual's credit/debit card, everyone involved in the process suffers,
right from the individual whose confidential data has been leaked to the
businesses (generally banks) who issue the credit card and the merchant
who is finalizing the transaction with purchase. This makes it extremely
essential to identify the fraudulent transactions at the onset. Financial
institutions and businesses like e-commerce are taking firm steps to flag
the fraudsters entering the system. Various advanced machine learning
technologies are at play, assessing every transaction and stemming the
fraud users in its nip using behavioural data and transaction patterns. The
process of automatically differentiating between fraudulent and genuine
users is known as “credit card fraud detection.”

1|Page
3 HOW DOES CREDIT CARD FRAUD WORK?

A credit card is one of the most used financial products to make online
purchases and payments such as gas, groceries, TVs, traveling, shopping
bills, and so on because of the non-availability of funds at that instance.
Credit cards are of most value that provide various benefits in the form
of points while using them for different transactions. There are several
categories of credit card fraud that are prevalent in today’s time:

• Lost/Stolen cards: People steal credit cards from the mail and
use them illegally on behalf of the owner. The process of blocking
credit cards that have been stolen and re-issuing them is a hassle
for both customers and credit card companies. Some financial
institutions keep the credit cards blocked until it is verified that the
rightful owner has received the card.
• Card Abuse: The customer buys goods and items on the credit
card but has no intention to pay back the amount charged by the
bank for the same. These customers stop answering the calls as the
deadline to settle the dues approaches. Sometimes they even
declare bankruptcy—this type of fraud results in losses of millions
every year.
• Identity Theft: The customers apply illegitimate information,
and they might even steal the details of a genuine customer to
apply for a credit card and then misuse it. In such cases, even card
blocking cannot stop the credit card from falling into the wrong
hands.
• Merchant Abuse: Some merchants show illegal transactions
(that never occurred) for money laundering. For performing these
illicit transactions, legal information of genuine credit card users is
stolen to generate replicas of the cards and use it for illegal work.

Financial institutions increasingly depend upon automated machine

learning systems to make intelligent decisions and protect businesses
against substantial losses. These measures play a significant role in
reducing the risk while doing online transactions. Like humans, machine
learning algorithms learn from past transaction data and use that
2|Page
information to analyse future transactions with the same lens. While
machines might not be as intelligent as humans are and might need
some supervision on top of it, the advantage lies in the speed of data
processing and computation. Also, machines can identify and remember
more patterns in vast volumes of data compared to humans.

2 OBJECTIVE AND SCOPE

➢ When fraud occurs, the cardholder disputes the accuse and the
deduction is usually cancelled, which means either the cardholder's
bank or the mercantile absorbs the loss.
➢ Due to the rise and hasty expansion of E-Commerce, use of credit
cards for online purchase has considerably increased and it caused
an explosion in the credit card fraud.
➢ As credit card becomes the majority well-liked mode of payment
for both online as well as usual purchase, cases of fraud related
with it are also growing
➢ In real life, fraudulent transactions are spotted with actual
transactions and simple pattern matching technique are not often
enough to detect those frauds perfectly.
➢ Performance of efficient fraud recognition systems has thus
become imperative for all credit card issuing banks to reduce their
losses.

2.1 NEED FOR THE PROJECT

➢ The reason may be to obtain merchandise without paying, or to get

unauthorized rites from an account. Credit card fraud is also an
addition to detect the thief.
➢ Immediate detection of fraudulent activities.

3|Page
2.2 PROBLEM IDENTIFICATION

➢ Two vast data sets of real-world transactions show that, in order to

get precise alerts, it is mandatory to assign larger importance to
feedbacks the learning problem.
➢ Not unexpectedly, feedbacks play a mid-role in the planned
learning approach, which consists in separately.
➢ Classifiers that blend feedbacks and postponed supervised sample
or that implement case weighting scheme are often returning less
precise alerts.
➢ A person's information always treated as a fraud if it once detected
as a fraud which reduces flexibility of the syste.

3 EXISTING SYSTEM

➢ Fraud is detected after the fraud is done, and the fraud is detected
after the complaint of the cardholder.
➢ The transaction is maintained in a record, we need to uphold a vast
data.
➢ Fraud is detected in a superlative and easy way.
➢ A distributed cost-based model is presented in that technique a
parallel and data learning technique is presented which provides
an efficient technique for fraud detection.

3.1 PROPOSED SYSTEM

➢ Using Clustering and Classification Technique we can find the
fraud transaction and original transaction.
➢ The detection can be done automatically so it will detect the fraud
while transaction.
➢ The accuracy of the detection is more accurate by using the
clustering techniques.
➢ It will predict using features of the user already transacted details
cording to the purchase and card details, Semi Supervised
approach is used to detect the fraudulent.

4|Page
3.1 ADVANTAGES OF PROPOSED SYSTEM

➢ Will not mark any user so misunderstanding of user and fraud is

avoided.
➢ Accuracy is better than the existing system.
➢ Semi supervised approach is used so that no need of all feature
should need the labels.
➢ Historical data is used so we can get the pattern of user's original
transaction and Fraud transaction.

4 REQUIREMENT SPECIFICATION

Hardware Requirements

Processor Pentium Dual Core 2.3 GHz

Hard Disk 250 GB or Higher

RAM 2 GB (Min)

Software Requirements

Operating System Windows 7 or Higher

Languages used Python( Pandas, NumPy, Sklearn)

Tools Anaconda, Jupyter Notebook,

Spyder

5|Page