Journal of Science and Technology

ISSN: 2456-5660 Volume 4, Issue 06 (NOV-DEC2019)

www.jst.org.in DOI:https://doi.org/10.46243/jst.2019.v4.i06.pp37-41
DETECTION OF CREDIT CARD FRAUD WITH RANDOM
FOREST ALOGIRITHAM
Dr.D.Sharada Mani1, Mrs.M.Lakshmi Tirupathamma2, Mr.M.Hanumantha Rao3, Mrs.B.Sujatha4

Professor1, Assistant Professor2,3,4

Department of CSE
NRI INSTITUTE OF TECHNOLOGY, Visadala X Road, Medikonduru(M), Guntur (Dist), Andhra Pradesh 522438.

Abstract
Real-world credit card fraud detection is the primary emphasis of the project. Credit card fraud has lately increased dramatically as a result of
the amazing surge in the number of transactions. The goal is to get something without paying for it or to get money out of a bank account
without authorization. All credit card issuers must now have effective fraud detection systems in order to reduce their losses. Making the
business is a major difficulty since no cardholder or card must be present for a transaction to be completed.. Merchants are unable to
determine if a consumer presenting their card is in fact the legitimate owner. It is possible to increase the accuracy of fraud detection by using
the suggested technique, which makes use of a random forest. The random forest technique is used to analyse a data set and the current
dataset of the user. Finally, improve the precision of the output data. The accuracy, sensitivity, specificity, and precision of the procedures are
assessed. Processed characteristics are used to identify fraud, and a graphical model depiction is presented. The accuracy, sensitivity,
specificity, and precision of the procedures are assessed.

Keywords:Fraud Detection, Random Forest, Credit Card.

INTRODUCTION
In credit card transactions, researchers have kept in mind a variety of methodologies to construct models based on
artificial intelligence, data mining, fuzzy logic, and machine learning in order to identify fraudulent activity.
Detecting credit card fraud is a challenging, but common, challenge. Machine learning was used to develop the
credit card fraud detection in our suggested system. Machine learning is becoming more sophisticated. Machine
learning has been shown to be effective at detecting fraud. During online transactions, a significant quantity of data
is exchanged, resulting in a binary outcome: legitimate or fraudulent. Features are built into the bogus datasets.
Customers' age and credit card balance are only two examples of these data pieces. There are hundreds of
characteristics, and each one has a different impact on the likelihood of a transaction being fraudulent. Note that the
machine's artificial intelligence, which is guided by the training set, generates the level at which each characteristic
contributes to the fraud score. This level is not chosen by a fraud analyst. Accordingly, a transaction that is made
using a credit card will have the same fraud weighting as one that is made using a debit card. But if this decreased,
the contribution level would also decrease at the same rate. These models are self-taught and do not need any further
programming, such as a manual review. Classification and regression methods are used for credit card fraud
detection in Machine learning. For online or offline fraud card transactions, we employ a supervised learning system
such as the Random Forest method. An evolved variant of Decision Trees is Random Forest. Random forest
outperforms all other machine learning algorithms in terms of performance and accuracy. A key goal of random
forest is to alleviate the problem of feature space oversampling, which was discussed before. De-correlated trees are
pruned by setting a stopping point for node splits, which I'll explore in more depth later on in this article.

PROBLEM DEFINITION
Fraudulent credit card transactions result in billions of dollars in losses each year. As ancient as mankind, deception
may take on almost any shape. According to the 2017 PwC global economic crime study, around 48% of firms have
been the victim of economic crime. As a result, there is a pressing need to find a solution to the credit card fraud
detection issue. As new technologies emerge, fraudsters now have more options than ever before. Using a credit
card is common in today's culture, and the number of people falling victim to credit card fraud is on the rise. Hugh
In addition to merchants and banks, those who use credit cards are affected by financial losses that have been
fraudulently perpetrated. For a business, non-financial costs from fraud might be difficult to measure in the near
term, but they may become apparent over time. It's very uncommon for customers to switch credit card providers
after becoming a victim of fraud with a particular organisation.

SCOPE OF THE PROJECT

Published by: Longman Publishers www.jst.org.in

Journal of Science and Technology
ISSN: 2456-5660 Volume 4, Issue 06 (NOV-DEC2019)
www.jst.org.in DOI:https://doi.org/10.46243/jst.2019.v4.i06.pp37-41
Keeping track of fraudulent transactions is becoming more challenging as new technology is introduced. It is now
possible to automate and save part of the effective work that goes into identifying credit card fraud thanks to the
growth of machine learning, artificial intelligence, and other key information technology sectors.

RELATED WORK
[1] "KosemaniTemitayo Hafiz, Dr. Shaun Aghili, Dr. PavolZavarsky" use predictive analytics technology to detect
credit card fraud in Canada.
Predictive analytics vendor solutions presently utilised to identify credit card fraud are the subject of this paper's
assessment criteria, features, and capabilities. The scorecard compares five credit card predictive analytics vendor
solutions that have been implemented in Canada against each other. A list of credit card fraud PAT vendor solution
problems, hazards, and limits was compiled as a result of the study results.
Card fraud may be detected with the use of hybrid methods like [2] BLAST-SSAHA. The authors of the paper are
"AmlanKundu, SuvasiniPanigrahi, Shamik Sural, Senior Member of the IEEE, and Arun K. Majumdar" Paper
proposes a two-stage sequence alignment in which the profile Analyzer (PA) first compares incoming transaction
sequences to the real cardholder's prior spending sequences, and then compares these two sets of transactions. The
profile analyzer then sends the strange transactions it found to a deviation analyzer (DA) to see whether they match
up with previous fraudulent behaviour. It is up to these two analysts to make the ultimate determination on the
nature of a transaction. Combining two sequence alignment methods such as BLAST and SSAHA in order to attain
online response times of both PA and DA is suggested.Researchers Investigate a Distance Sum Detection Model for
Credit Card Fraud"Wen-Fang YU, Na Wang," says the voiceover. Due to the rapid expansion of China's credit card
market and increased amount of international commerce, there has been an uptick in credit card fraud. The emphasis
of bank risk management is now on how to better identify and prevent credit card fraud. Outlier detection based on
distance sum is proposed as a fraud detection model for credit card transactions that takes into account both the
rarity and unusual nature of credit card fraud. Using this approach to identify credit card fraud is practical and
accurate, according to the results of several experiments.Detection of Fraudulent Credit Card Transactions Using
SVM and Decision Trees. Nipane, Poonam, Kalinge, Vidhate, Vidhate, Kunal, and Deshpande are the members of
this group.Fraud is rising over the globe as e-commerce advances, resulting in massive losses for businesses.
Currently, credit card fraud is a major source of financial loss for both businesses and consumers. They include
decision trees, genetic algorithms, meta-learning strategies and neural networks, as well as HMMs. Support Vector
Machine (SVM) and decision trees are being employed to tackle the challenge in a contemplated system for fraud
detection. Financial losses may be minimised to a larger extent by the use of a hybrid method..Credit Card Fraud
Detection using Supervised Machine Learning (SVM). SITARAM PATEL AND SUNITA GOND.There are
numerous kernels in the SVM (Support Vector Machine) based approach proposed in this thesis, instead of only the
spending profile. To summarise, there has been significant progress made in the simulation in terms of the rates of
both "true positive" and "true negative," as well as reductions in the rates of "false positives" and "false
negatives."Support Vector Machines and Decision Trees for Credit Card Fraud Detection, by Y. Sahin and E.
DumanIn this work, decision trees and support vector machines (SVM) classification models are constructed and
used to the challenge of detecting credit card fraud. SVM and decision tree algorithms were compared in this work
for the first time using an actual data set for credit card fraud detection.

SYSTEM ANALYSIS
EXISTING SYSTEM
Cluster Analysis and Artificial Neural Networks have been used to identify credit card fraud in an existing system,
where data normalisation is conducted before to Cluster Analysis, and findings demonstrate that by grouping
characteristics, neuronal inputs may be decreased. Using normalised data with MLP-trained data may provide good
results. Unsupervised learning was used in this study. Finding novel approaches for fraud detection and increasing
the accuracy of outcomes were the main goals of this article. Real-world transactions from a big European
corporation are used in this study, and personal information is kept private. An algorithm's accuracy hovers around
50%. The purpose of this work was to identify an algorithm and lower the cost metric in some way. By 23%, they
were able to come up with Bayes' lowest risk method.

Disadvantages

Published by: Longman Publishers www.jst.org.in

Journal of Science and Technology
ISSN: 2456-5660 Volume 4, Issue 06 (NOV-DEC2019)
www.jst.org.in DOI:https://doi.org/10.46243/jst.2019.v4.i06.pp37-41
One of the novel collative comparison measures provided in this study appropriately depicts the profits and losses
owing to fraud detection. The suggested cost measure is used to demonstrate a Bayes minimal risk technique that is
cost sensitive.

PROPOSED SCHEME
To classify credit card data, we are using a random forest algorithm in our suggested system. Classification and
regression are both possible with Random Forest. Basically, it's a set of decision tree classifiers. Over-fitting is a
problem for decision trees, although random forests solve this problem. Randomly sampling a portion of the training
set is used to train each tree in the decision tree, which then divides on a feature taken from the whole feature set.
For big data sets with numerous characteristics and instances, training in a random forest is incredibly rapid since
each tree is trained independently of the others. For example, the Random Forest approach provides a decent
estimate of generalisation error and is tolerant to over fitting.

ADVANTAGES OFPROPOSED SYSTEM

In a regression or classification task, Random Forest can naturally rank the relevance of variables. • The 'amount'
feature refers to the total transaction value. Features 'class' are used in the binary classification and take values of 1
(fraud) or 0 (no fraud), depending on the outcome of the classification (not fraud).

SYSTEM ARCHITECTURE
Cleansing and validation are conducted on credit card data before it is separated into two parts: one for training and
the other for testing. The training dataset is used to train the model, while testing datasets are used to evaluate the
model's performance. Teat and train datasets have now been generated from the original sample at random.

Figure 6.1 shows the proposed system's architecture.

Algorithm Utilized
Random Forest
Machine learning algorithm Random Forest relies on the concept of ensemble learning. Combining several
algorithms or using the same technique numerous times to build a more effective prediction model is known as
ensemble learning. Multiple decision trees are combined in the Random Forest method, resulting in a forest of trees,
which is why the term "Random Forest" was coined. For both regression and classification, the random forest
technique is a good choice.

WORKING OF RANDOM FOREST

In order to run the random forest algorithm, you must follow these fundamental steps: To begin, choose N records at
random from the dataset. Create a decision tree based on the N data that you've collected so far. Repeat steps 1 and 2
for the number of trees you wish to use in your algorithm. Each tree in the forest has the ability to forecast the
category to which a new record will belong in the classification issue. Finally, the new record is awarded to the
category that receives the most votes, and therefore the new record.

ADVANTAGES OF USING RANDOM FOREST

Published by: Longman Publishers www.jst.org.in

Journal of Science and Technology
ISSN: 2456-5660 Volume 4, Issue 06 (NOV-DEC2019)
www.jst.org.in DOI:https://doi.org/10.46243/jst.2019.v4.i06.pp37-41
The advantages of classifying and predicting using random forests. Since there are several trees and each tree is
trained on a portion of data, the random forest approach is not biassed. The random forest method depends on the
strength of "the crowd," which reduces the system's overall slant. There is a high degree of predictability in this
method. However, the overall method is not much impacted by new data since even if one tree is impacted by the
new data, it is very unlikely to influence all trees. With a combination of category and numerical variables, the
random forest method performs well. Data with missing values or that has not been scaled adequately might also
benefit from the random forest approach because of its robustness.

CONCLUSION
More training data improves Random Forest's performance, but testing and application speed decrease. More pre-
processing procedures might also be beneficial. To get better results from SVM, additional preprocessing is needed
on the data, which is still an issue for the SVM method, despite the fact that the results are excellent.

REFERENCES
[1] Sudhamathy G: Credit Risk Analysis and Prediction Modelling of Bank Loans Using R, vol. 8, no-5, pp. 1954-1966.

[2] LI Changjian, HU Peng: Credit Risk Assessment for ural Credit Cooperatives based on Improved Neural Network, International
Conference on Smart Grid and Electrical Automation vol. 60, no. - 3, pp 227-230, 2017.

[3] Wei Sun, Chen-Guang Yang, Jian-Xun Qi: Credit Risk Assessment in Commercial Banks Based On Support Vector Machines, vol.6,
pp 2430-2433, 2006.

[4] AmlanKundu, SuvasiniPanigrahi, Shamik Sural, Senior Member, IEEE,“BLAST-SSAHA Hybridization for Credit Card Fraud
Detection”, vol. 6, no. 4 pp. 309-315, 2009.

[5] Y. Sahin and E. Duman, “Detecting Credit Card Fraud by Decision Trees and Support Vector Machines, Proceedings of
International Multi Conference of Engineers and Computer Scientists, vol. I, 2011

[6] Sitarampatel, SunitaGond , “Supervised Machine (SVM) Learning for Credit Card Fraud Detection, International of engineering
trends and technology, vol. 8, no. -3, pp. 137- 140, 2014.

[7] SnehalPatil, HarshadaSomavanshi, JyotiGaikwad, AmrutaDeshmane, RinkuBadgujar," Credit Card Fraud Detection Using Decision
Tree Induction Algorithm, International Journal of Computer Science and Mobile Computing, Vol.4 Issue.4, April- 2015, pg. 92-95 [

8] Dahee Choi and Kyungho Lee, “Machine Learning based Approach to Financial Fraud Detection Process in Mobile Payment
System", vol. 5, no. - 4, December 2017, pp. 12-24

Published by: Longman Publishers www.jst.org.in