Scribd
Upload
25 views3 pages

Detecting Fraudulent Transactions

The document discusses the development of models for detecting credit card fraud using various machine learning algorithms, including Random Forest, Logistic Regression, Support Vector Machine (SVM), and Naïve Bayes. It highlights the importance of extracting relevant features from transactional datasets to improve detection accuracy. The study compares the accuracy rates of these algorithms, concluding that Random Forest achieved the highest accuracy at 98.60%.

Uploaded by

blossomvrinda
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
25 views3 pages

Detecting Fraudulent Transactions

The document discusses the development of models for detecting credit card fraud using various machine learning algorithms, including Random Forest, Logistic Regression, Support Vector Machine (SVM), and Naïve Bayes. It highlights the importance of extracting relevant features from transactional datasets to improve detection accuracy. The study compares the accuracy rates of these algorithms, concluding that Random Forest achieved the highest accuracy at 98.60%.

Uploaded by

blossomvrinda
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT (IJSREM)

VOLUME: 04 ISSUE: 05 | MAY -2020 ISSN: 2582-3930

DETECTING FRAUDULENT TRANSACTIONS

Tanoo Bijapur Sushma Managundi Anusha Shetty


dept. Of Information Science dept. Of Information Science dept. Of Information Science
S.D.M College of S.D.M College of S.D.M College of
engineering engineering engineering
Dharwad,India Dharwad,India Dharwad,India

Under the guidance of ,


Lydia Adina
dept. Of Information Science Dr.V K Parvati
S.D.M College of
dept. Of Information Science
engineering
S.D.M College of
Dharwad,India
engineering
Dharwad,India

Abstract—When constructing a credit card fraud models are built which help in detecting fraud or
detection model we extract the right features from not.
the transactional dataset. In this paper we propose Fraud can be defined as the intent of deception in
to create the respective models for Random forest, order to obtain financial gain and without the
Logistic regression, Support vector machine(SVM) knowledge of the card owner and the Bank.
and Naïve Bayes algorithm. And find the accuracy Credit card Fraud can be of types:
rates of each algorithms and get the most accurate 1. CARD-NOT PRESENT(CNP)
algorithm to detect fraudulent transaction.
2. LOST AND STOLEN CARDS
3. CARD ID THEFT
I. INTRODUCTION 4. ACCOUNT TAKEOVER.
Due to advancement in E-commerce, the online
transactions have increased which has caused
increase in the fraudulent transactions.
In the era of digital world it is necessary to identify
II.ALGORITHMS
the frauds and provide appropriate solutions by
understanding various technologies algorithms and 1.Random Forest
types involved in detection. Forest is a collection of decision trees.Decision tree
The dataset of transactions is given to the machine, are mainly indulged in overfitting of training
which understands the data by analysing. And data.They show high variation for small change in
using machine learning algorithms the respective the input data,hence is sensitive to specific
data.The input data can be trained to ensure that it

© 2020, IJSREM | www.ijsrem.com Page 1


INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT (IJSREM)
VOLUME: 04 ISSUE: 05 | MAY -2020 ISSN: 2582-3930

remains error-proned.This algorithm helps to grow III. COMPARISION BETWEEN THE


many such decision trees and gives average of all ALGORITHMS
classification trees,hence it results in the reduction
of variance.The random forest model is built to ALGORI ACCURAC ADVANT DISADV
which a specified dataset is given,it gives the THM Y RATE AGES ANTAGE
appropriate resultant value. S
Random 98.60% Predictive Training
2.SVM forest performa a large
ML feed the data to machines and then it learns nce can number
from the data without explicitly programming it.It compete of deep
can be used for both classification and regression with the trees can
algorithms. It uses two hyperplane to separate two best have high
new classes and then check the distance between supervise computat
them. d ional cost
learning s
3.Naïve Bayes algorithm
Naïve Bayes is based on the Assumption which has s
independence amongst the Predictors. In simple Logistic 97.50% Performs Assumptio
terms, this could be put up as Naive Bayes Regressio well when n of
Classifier which assumes that a particular feature in n dataset is linearity of
a class is not exactly directly related to any other linearly dataset
feature. separable. b/w
P(H/E) = P(E/H) / P(E) independe
Naive Bayes model isn’t difficult to build and is nt and
really useful for very large datasets. Along with dependent
simplicity, Naive Bayes is also considered to have variable.
outperformed all the highly sophisticated Naïve 90.00% Its easily It assumes
classification methods. Bayes trained every
with small feature is
4.Logistic Regression dataset
A logistic regression model is termed as a independe
probabilistic model. It helps in finding the nt which is
probability that a new instance not true
always.
Support 97.70% It is It doesn’t
belongs to a certain class. Since it is probability, vector effective provide
the output lies between 0 and 1. machine in high probability
Whenever we are using the logistic regression as a dimension estimates.
binary classifier, we can consider the classes to be al spaces
a positive class and a negative class. We then find
the probability. Higher the probability (greater than
0.5), it is likelier that it falls into the positive class. IV. CONCLUSION
Similarly, if the probability is low (less than0.5), In this paper we try to find whether its fraud or not
we can classify this into the negative class. using the algorithms. We calculate the accuracy
rates of each algorithms and by comparing the
accuracy rates obtained, the best algorithm be
determined.

© 2020, IJSREM | www.ijsrem.com Page 2


INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT (IJSREM)
VOLUME: 04 ISSUE: 05 | MAY -2020 ISSN: 2582-3930

We extract the essential features from the [3] Karunya University “Analysis on credit card fraud detection methods”
Asst professor Annie Portia.
transactional data and try to find the genuine [4] IEEE journal,”A novel approach using aggregation strategy and
feature from all these algorithms. feedback mechanism” Changun jiang,2017
[5] Tongji University,”Random forest for credit card fraud
REFERENCES detection”,Shiyang xuon ,Shangai China.
[6] Tongji University,”Algorithms for detection”Guanjum liu,China.

[1] University of lexembourg,”Deteceting fraud detection using periodic


features”Aljendro Correa Bahnsen.
[2] Institute of Control Sciences” credit card fraud Detection
system”,Moscow Russia

© 2020, IJSREM | www.ijsrem.com Page 3

You might also like