SRM INSTITUTE OF SCIENCE AND TECHNOLOGY

Ramapuram , Chennai – 600 089

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

18CSP106L-SEMINAR - 2
FRAUD DETECTION USING ML
BATCH NUMBER :15
Team Members Supervisor
Shaurya Jain NAME:
Aditya Pund
Tejas Kumar Dev

Date Slide Number

 Agenda(Only for Reference-Modify
accordingly)
• Abstract
• Scope and Motivation
• Introduction
• Literature Survey ( Table)
• Objective
• Problem Statement
• Proposed Work
• Architecture Diagram/Flow Diagram/Block Diagram
• Novel idea
• Modules
• Module Description
• Software & Hardware Requirements
• Implementation (Complete Demo to be shown)
• Results and Discussion
• Conclusion
• Future Work
• References (Min. 15 references)
• Outcome ( Paper Submitted Proof /Patent Filing Proof)

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

 ABSTRACT
• A state-of-the-art method of protecting digital financial
transactions is to use machine learning for online transaction
fraud detection. Complex algorithms are utilized in this process
to assess the vast amount of data collected during online
payments in order to detect potentially fraudulent conduct.
These algorithms monitor user behaviour, transaction history,
and other data to spot any suspicious patterns, such as strange
purchase locations or atypical spending amounts. The
technology can block transactions automatically or even send
out alerts when it identifies probable fraud in order to avoid
unwanted access. This technology has grown more and more
crucial in the digital age, as an increasing number of banking and
e-commerce transactions are completed online.
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING
 Scope and Motivation
Image segmentation and in painting are both essential tasks in computer vision
and image processing, with various applications and promising research areas.
• Financial Institutions:
Banks and credit card companies employ machine learning-based fraud
detection to protect their customers from unauthorized transactions. It helps
detect credit card fraud, account takeovers, and other financial crimes in real-
time.
• E-commerce Platforms:
Online retailers use fraud detection systems to identify and prevent fraudulent
purchases. This safeguards both the businesses and their customers by
reducing charge backs and losses due to fraudulent orders.
• Payment Service Provider:
Companies offering digital payment solutions leverage machine learning to
monitor transactions and prevent fraudulent activities, enhancing the security
of online payments.
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING
 Introduction
When you make purchases or transact on the internet, machine learning-
powered online transaction fraud detection acts as a kind of digital guardian,
safeguarding your money and personal data. Consider use your credit card to
make purchases online or in-store. Behind the scenes, a lot goes on, and not
everything is sincere. Some attempt to deceive the system by fabricating
transactions or obtaining your personal financial information. Herein lies the
role of machine learning. It's like to having an extremely intelligent robot
watching every transaction all the time, picking up on patterns and identifying
any odd or suspect conduct. It considers factors such as your location,
purchasing patterns, and typical shopping behaviors. In our increasingly online
society, the significance of online transaction fraud detection has increased
dramatically. As more and more individuals shop, bank, and make investments
online, fraud chances have increased as well. The ever- evolving tactics of
hackers and scammers necessitate the need for an intelligent system that can
adapt and stay one step ahead. This is made feasible by machine learning,
which can adapt to the shifting strategies used by con artists.
DEPARTMENT OF COMPUTER SCIENCE AND
ENGINEERING
 Literature Survey
S.No. Title of the Year Journal/ Inferences
Paper Conference
Name

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

 Objective
This project's main goal is to use machine learning to create and
implement an effective system for detecting online transaction fraud.
Through the analysis of numerous data points and trends, this system
seeks to safeguard consumers and financial institutions from fraudulent
activities in online transactions by identifying and preventing fraudulent
transactions. The primary challenges to address in this project are:

• Evolving Fraud Tactics:

• Data Volume and Complexity:
• Efficiency and Real-Time Capability:
• Reducing False Positives:

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

 Problem Statement
While creating an Online Transaction Fraud Detection
using Machine Learning, several challenges and problems
may arise. Some of the key issues include:
• Increasing online fraud incidents.
• Ineffectiveness of traditional rule-based systems.
• Need for real-time fraud detection.
• Complexity and volume of transaction data.
• Requirement for machine learning solutions.
• Ensuring user confidence in online transactions.
• Adapting to evolving fraud tactics.
• Minimizing financial losses and identity theft.
DEPARTMENT OF COMPUTER SCIENCE AND
ENGINEERING
 Proposed Work
One highly effective model for online transaction fraud detection using machine
learning is the Gradient Boosting Machine (GBM). GBM is a supervised learning
technique that combines multiple weak predictive models to create a powerful
ensemble model. It operates by building decision trees sequentially, each tree
learning from the errors of its predecessor has become a popular choice due to its
ability to handle large datasets, nonlinear relationships, and imbalanced data—a
common characteristic in fraud detection where the number of fraudulent cases is
significantly lower than legitimate ones. It excels in detecting anomalies, as it's
capable of focusing on the misclassified data points, thus enhancing model
accuracy with each iteration. This model's adaptability and robustness against
imbalanced data, coupled with its high predictive accuracy and efficiency, make it
a leading choice in the field of fraud detection. Organizations frequently deploy
GBM in their fraud detection systems to continuously learn and adapt to new
fraud patterns, ensuring a more secure and trustworthy digital financial
ecosystem. DEPARTMENT OF COMPUTER SCIENCE AND
ENGINEERING
Software & Hardware Requirements

DEPARTMENT OF COMPUTER SCIENCE AND

ENGINEERING
Implementation

DEPARTMENT OF COMPUTER SCIENCE AND

ENGINEERING
 Results and Discussion
• Screenshots of Output
• Graphs
• Tabulated Results

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

 Conclusion

DEPARTMENT OF COMPUTER SCIENCE AND

ENGINEERING
Future Work

DEPARTMENT OF COMPUTER SCIENCE AND

ENGINEERING
 References

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

 Outcome
• Proof for Paper Submission/Patent Filing
• Research Paper

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING