Credit Card Fraud Detection & Prevention

of Fraud Using Genetic Algorithm

Ashish Sinha (22051846) Ayush Aryan (22051847) Devanshu Pundir (22051853)
School of Computer Engineering School of Computer Engineering School of Computer Engineering
KIIT Deemed to be University KIIT Deemed to be University KIIT Deemed to be University
Bhubaneshwar, Odisha, India Bhubaneshwar, Odisha, India Bhubaneshwar, Odisha, India
[email protected] [email protected] [email protected]

Abstract—Companies and institutions are shifting parts of their card transactions. The measure is necessary because the
operations, or even their entire business, to online platforms. This manuscript was received in January 2013.Tech. Add twotag
move allows them to offer e-commerce, information, and where required for a line break. Rinky D. Patel,Dheeraj
communication services to their customers, enhancing efficiency
Kumar Singh,Gujarat Technological University, Gujarat,
and accessibility. Payment card fraud has emerged as a
significant issue globally. Companies and institutions incur
IndiaInformation Technology,Information Technology,Parul
immense financial losses each year as a result of fraud. Institute of Engg. Please insert a line break where necessary:
Fraudsters persistently explore innovative methods to engage in Parul. Institute ofEngg. The researcher holds a Ph. D. in
unlawful activities. In this assessment, our focus will be on Science and Technology from Gujarat Technological
identifying fraudulent transactions utilizing the genetic University, located in Gujarat, India. The inherent structure
algorithm.Genetic algorithms are utilized to determine the of credit card (CC) transactions should be examined
network topology, the number of hidden layers, and the quantity closely.This isabout enhancing the parametric fraud
of nodes required for designing the neural network for our credit detection solution. The level of losses resulting from fraud,
card fraud detection problem.
along with the understanding of the correlation between
Keywords-Communication,financial losses and the credit card's available limit, has prompted us
loss,fraudster,transaction,genetic algorithm,assesment to devise an effective performance solution. In recent years,
the primary focus has been on data mining in the context of
developing a model for credit card fraud detection. Since our
I. INTRODUCTION issue is viewed as a classification problem, traditional data
mining algorithms are not suitable. An alternative approach
The credit card is a small plastic card issued to users as a can be pursued employing general-purpose heuristic
system of payment. It permits its cardholder to purchase methods such as genetic algorithms.This paper is aimed at
goods and services in exchange for the cardholder's proposing a credit card fraud detection system using genetic
commitment to settle the payment for them. Credit card algorithm. Genetic algorithms are evolutionary algorithms
security is dependent on the physical security of the plastic that strive to improve solutions over time.When a card is
card, along with the privacy of the credit card number. duplicated, stolen, or lost and taken by fraudsters, it is
Globalization and the growing popularity of online typically used until its available limit is exhausted.Instead of
shopping, along with the increased use of the internet, have focusing solely on the count of accurately classified
led to a significant rise in credit card transactions transactions, a more notable approach is to minimize the
worldwide.A quick increase in the volume of credit card overall available limit on cards affected by fraud.The goal is
transactions has resulted in a significant surge in fraudulent to reduce false alerts by utilizing a genetic algorithm to
activities.Credit card fraud covers a broad spectrum of theft optimize a set of interval-valued parameters. Among
and fraudulent activities carried out by utilizing a credit card decision trees, they are more popular. Fraud detection
as an illegitimate source of funds during a transaction. typically falls within the domain of E-commerce and data
Credit card fraudsters use a variety of methods to carry out mining. The Genetic algorithms are evolutionary algorithms
fraudulent activities.To effectively combat credit card fraud, in which the primary goal is to achieve superior solutions. It
it is crucial to first grasp the mechanisms for identifying is crucial, particularly in combatting fraud, to prioritize the
such fraudulent activities. Over time, credit card fraud has development of efficient and secure electronic payment
become more stable thanks to the implementation of various systems. These systems play a key role in determining the
fraud detection and prevention mechanisms. Fraud can be legitimacy of transactions by detecting and preventing
described as the undesired activities occurring within an fraudulent activities.
operational system. Fraud can manifest in a diverse range of
areas, such as finance, telecommunications, health care, and
public services. Let's delve into the topic of Credit Card
II. LITERATURE REVIEW

Fraud detection requires keeping an eye on users' actions to

Fraud. Credit Card Fraud occurs when one person uses predict, identify, or prevent any unwanted behaviors.To
another individual's credit card for personal gain without the effectively combat credit card fraud, it's essential to grasp
knowledge of the card owner or issuer. This unauthorized the technologies used in detecting fraudulent activities and
usage involves accessing an account that was not meant for to recognize the different types of credit card
their use. In this study, we are exploring financial frauds and frauds.Numerous algorithms are available for the detection
will specifically concentrate on identifying fraudulent credit of credit card fraud.They are artificial neural-network
models based on artificial intelligence and machine learning
approach, distributed data mining systems, sequence minimized.The detection of fraud must occur in real-time,
alignment algorithm based on the spending profile of the while also ensuring that the frequency of false alerts is kept
cardholder, intelligent decision engines based on artificial to a minimum.There are various devices available to assist
intelligence, Meta learning Agents, and Fuzzy based with conducting that transaction.The potential actions
systems. The use of various technologies in detecting credit include blocking the card,sending an SMS, or contacting the
card fraud includes a Web Services-Based Collaborative cardholder for a call. In financial institutions, fraud detection
Scheme for Credit Card Fraud Detection, where participant is implemented by analyzing customer behavior
banks can share information on fraud patterns to strengthen variables.The sample data set is used to generate fraudulent
their detection capabilities and minimize financial losses. transactions and detect fraud in electronic payment systems.
Other methods involve Artificial Immune System for Credit The data set includes various parameters.
Card Fraud Detection, CARDWATCH utilizing Neural
Network for Database Mining, Adaptive Fraud Detection CCfreq= number of times card used
and Prevention with Genetic Algorithm, Real-time fraud
CCloc = location at which CCs in the hands of fraudsters
detection using computational intelligence, and credit card
fraud detection through self-organizing maps. Many credit CCoverdraft = the rate of overdraft time
card fraud detection systems utilize artificial intelligence,
meta-learning, and pattern matching. The aim is to create a CCbank balance = the balance available at bank of CC
credit card fraud detection system utilizing genetic CCdailyspending = the average daily spending amount
algorithm.During the credit card transaction, fraud is
detected. The number of false alerts is minimized through Data set T={t1,t2,t3,…. ,tn}, U is one data object, If p parts
the use of a genetic algorithm.Instead of focusing solely on of data set named S in data set is far away from object U, S
maximizing the number of correctly classified transactions, Є T, U Є T, then U is Common object. The proposed system
we have crafted an objective function that considers variable overcomes the above mentioned issue in an efficient way.
misclassification costs. This approach ensures that correctly Using a genetic algorithm, fraud can be identified, false
classifying certain transactions, deemed more crucial, takes alerts minimized, and optimized results achieved.The
precedence over others. The significant losses resulting from detection of fraud relies on the behavior of the customer. A
fraudulent activities and the understanding of the correlation fresh classification problem is presented, featuring a variable
between these losses and the available limit must be misclassification cost.Here, a genetic algorithm has been
developed to optimize a set of interval-valued parameters.

III. GENETIC ALGORITHM Standardize the data, obtain the final sample, which contains
confidential information about the cardholder, and store it in
Genetic Algorithm (GA) is an optimization technique the dataset.
that mimics the natural evolution process. Individuals
possessing favorable traits for adapting to the Step2. Calculate the critical values. Determine the CC usage
environment are more likely to reproduce and survive. frequency count, the CC usage location, CC overdraft,
These advantageous individuals mate with each other, current bank balance, and average daily spending.
producing offspring with similar characteristics. This
way, favorable traits are passed on while unfavorable Step 3. Generate critical values found after a limited number
ones are eliminated, thus contributing to the gradual of generations.Critical Fraud DetectedMonitorable Fraud
evolution of the species. In simpler terms, the DetectedOrdinary Fraud Detected Utilizing genetic
fundamental concept behind genetic algorithms is that algorithms,
within a given problem, the genetic pool of a particular
population may hold the solution, or even an improved Step 4 is implemented.
solution. Drawing from genetic and evolutionary
principles, the genetic algorithm iteratively adjusts a Create fraudulent transactions using this algorithm.This
population of artificial structures by implementing analysis aims to evaluate the feasibility of credit card fraud
initialization, selection, crossover, and mutation detection using a technique that applies detection mining
operators to derive an evolved solution. The goal of an based on critical values. It introduces detection procedures
artificial genetic algorithm is to enhance problem- and outlines the process.
solving by maintaining the optimal combination of
input variables.It all begins with defining the Genetic algorithm The initial population is randomly
optimization problem. Next comes formulating an selected from a diverse sample space. The fitness value is
objective function to assess the potential candidate computed within each population and then organized
solutions (chromosomes). This function essentially accordingly. The selection process involves being chosen
determines the most effective individual. through a tournament method.

IV. RESULTS OR FINDING

The Experiment process consists of four steps.

Step 1: Input a group of data consisting of credit card

transactions. Each transaction record contains n attributes.
The Crossover is determined by employing a single point credit card transactions and frauds. In contemporary
probability.Mutation alters the new offspring through a situations, a method known as genetic algorithms is utilized
uniform probability measure.In the process of elitism as an alternative approach.This document suggests a fraud
selection, the finest solutions are carried forward to the next detection technique that utilizes genetic algorithms.The
generation. The new population is created and goes through genetic algorithms are evolutionary. Their primary goal is to
the same process. If the maximum number of generations is continuously improve solutions to effectively combat credit
reached, it is shown in Fig. card fraud. With this, we are able to create the most secure
payment system that can determine the legitimacy of our
The fundamental GA operators include crossover, selection, transactions.
and mutation. Selection, or survival of the fittest. The key to
making the right choice is to prioritize superior VI. REFERENCES
results.Mutation involves randomly trying combinations and [1] AGanesh Kumar.Nune, and P.Vasanth Sena, and T.P.Shekhar,
assessing the outcome, whether it is successful or ―Novel Artificial Neural Networks and Logistic Approach for
not.Crossover involves blending different positive elements Detecting Credit Card Deceit‖, In IJCSMR Vol 1 Issue 3 October
2012 ISSN 2278-733X October 2012.
to potentially generate an enhanced result. Please rewrite
[2] Raghavendra Patidar, and Lokesh Sharma, ―Credit Card Fraud
this text in a smoother tone. Include-where a line break is Detection Using Neural Network‖, In IJSCE ISSN: 2231-2307,
needed. Volume-1, Issue-NCAI2011, June 2011.
[3] Khyati Chaudhary, and Jyoti Yadav, and Bhawna Mallick, ―A
review of Fraud Detection Techniques: Credit Card‖, In IJCA Volume
45– No.1, May 2012.
[4] K.RamaKalyani, and D.UmaDevi, ― Fraud Detection of Credit Card
V. CONCLUSION AND FUTURE WORK Payment System by Genetic Algorithm‖, In IJSER Volume 3, Issue 7,
July 2012.
[5] Mehzabin Shaikh and Mrs. Gyankamal J. Chhajed, ― Review on
Financial Forecasting using Neural Network and Data Mining
Technique‖,In Global Journals Inc. Volume 12 Issue 11, 2012.
[6] Alejandro Correa, Banco Colpatria, Andres Gonzalez, Banco
Colpatria, camilo Ladino, Banco Colpatria, ―Genetic Algorithm
Optimization for Selecting the Best Architecture of a Multi-Layer
Perceptron Neural Network: A Credit Scoring Case‖ ,SAS Global
Forum, 2011.
[7] Khyati Chaudhary, Bhawna Mallick, ―Exploration of Data mining
techniques in Fraud Detection: Credit Card‖, In IJECSE ISSN 277-
1956/V1N3-1765-1771.
[8] Sushmita Mitra, Sankar K. Pal, Pabitra Mitra, ―Data Mining in Soft
Computing Framework: A Survey‖, IEEE Transactions On Neural
Networks, VOL. 13, NO. 1, January 2002.

Fig. :Process Flow of Genetic Algorithm

Credit card is a highly beneficial tool for making

payments.This helps the cardholders to purchase different
goods and services. This ensures that the cardholder will pay
for the goods and services they have purchased. The security
of a credit card relies on the confidentiality of its
number.With the widespread usage of credit cards in today's
world, we are encountering numerous instances of fraud.
Fraudsters employ various techniques to carry out credit
card fraud.Understanding concepts and identifying fraud
mechanisms empower us to effectively combat credit card
frauds. At present, there are numerous techniques at our
disposal to effectively combat credit card frauds.Credit Card
fraud occurs when an individual engages in unauthorized
activity on a credit card without the owner's knowledge.
Throughout this study, our main focus will be on detecting