Cost Sensitive Payment Fraud Detection Based On Dynamic Random Forest and KNN
Uploaded by
SafgbkjCost Sensitive Payment Fraud Detection Based On Dynamic Random Forest and KNN
Uploaded by
SafgbkjMALLA REDDY INSTITUTE OF TECHNOLOGY
EAMCET
AND SCIENCE Code: MRIT
An UGC Autonomous Institution JNTUH Code:
Approved by AICTE New Delhi and Affiliated to JNTU S1
An ISO 9001: 2015 Certified Institution
Accredited by NBA and NAAC ‘A’ Grade
Maisammaguda,
DEPARTMENT Medchal (Dist),SCIENCE
OF COMPUTER HyderabadENGINEERING
-500100, Telangana.
Email : [email protected], www.mrits.ac.in
Cost Sensitive Payment Fraud Detection Based ON
DynamiC Random Forest And KNN
PRESENTED BY
UNDER GUIDANCE
OF M.SRIKANTH
CH.CHANDHU
B.TECH,M.TECH
21S11A0509
ASSISTANT PROFESSOR
N.CHANDRA SEKHAR
21S11A0508
S.SACHIN
conten
t• Abstract
• Introduction
• Existing System
• Disadvantages
• Proposed Method
• Advantages
• System Architecture
• Hardware & Software Requirements
• Applications
• Result Analysis
• Future Scope
• Conclusion
• References
abstract
• Finance fraud is a growing problem with consequences in the financial
industry and data mining has been successfully applied to huge
volume of complex financial datasets to automate and analyze credit
card frauds in online transactions
• Data Mining is challenging process due to two major reasons–first,
profiles of normal and fraudulent behaviors change frequently and
second, card fraud data sets are highly skewed
• The results indicate about the optimal accuracy for Random Forest,
AdaBoost, XGBoost and LightGBM classifiers are 85%, 83%, 97.4%,
and 93% respectively.
INTRODUCTION
• The main focus has been given to recent computational methodologies
to handle the credit card fraud problem.
• Data mining technique is one notable and popular methods used in
solving credit fraud detection problem.
• We are using dynamic random forest and KNN for identifying the
froud in credit card transction.
EXISTING SYSTEM
• Existing systems utilize machine learning techniques like Support Vector
Machines (SVM) and Decision Trees, but struggle with high false positives.
• Others rely on manual rules and thresholds, leading to high false negatives
and difficulties in updating rules.
• Hybrid approaches combining machine learning and rule-based techniques
have also been used, but face challenges in integrating both methods and
adapting to new fraud patterns in real-time. Additionally, some systems
employ traditional Random Forest and K-Nearest Neighbors (KNN)
algorithms, but lack cost-sensitive learning and dynamic adaptability,
leading to limitations in accuracy and efficiency
disadvantages
• High false positives: Many systems flag legitimate transactions as
fraud, leading to unnecessary friction and customer frustration.
• High false negatives: Some systems miss actual fraud cases, resulting
in financial losses and damage to the business.
• Limited adaptability: Existing systems struggle to adapt to new fraud
patterns and schemes, making them less effective over time.
PROPOSED METHOD
• The proposed system utilizes a robust architecture that combines
advanced machine learning algorithms, including Dynamic Random
Forest and KNN, to analyze transactions from various angles.
• The system's real-time data processing capabilities allow it to
analyze transactions as they occur, leveraging user behavior analysis,
device fingerprinting, and transaction velocity to identify potential
fraud.
• The system's cost-sensitive learning approach ensures that it takes
into account the costs of false positives and false negatives,
optimizing its performance to minimize financial losses.
ADVANTAGES
•Improved Accuracy: The system's advanced algorithms and real-time processing
improve fraud detection accuracy, reducing financial losses.
•Adaptive Learning: The system updates its models in real-time, staying ahead of
evolving fraud patterns and improving detection capabilities.
•Real-Time Detection: The system detects fraud as it occurs, enabling swift action
to be taken and preventing potential losses.
System Architecture
REQUIREMENTS:
HARDWARE REQUIREMENTS:
System : Intel Core i3
64-bit, quad-core, 2.5 GHz minimum per core
Hard Disk : 1 GB
Ram : 4 GB
Display : Dual XGA (1024 x 768) or higher resolution
monitors
SOFTWARE REQUIREMENTS:
Operating system : Windows 10
Coding Language : Python 3.8.10
Library : TextBlob
APPLICATIONS:
• Financial institutions (banks, credit card companies)
• E-commerce platforms
• Insurance companies
• Government agencies (tax fraud, benefits fraud)
• Healthcare industry (medical billing fraud)
• Telecommunications companies
RESULT ANALYSIS:
User Interface
Upload Credit card Dataset
Generate Train & Test Model
Running Random Forest Algorithm
uploading Test Data
Detected Fraud From Test Data
Clean & Fraud Transaction Detection Graph
FUTURE SCOPE:
• Machine Learning and AI advancements: Integrating newer, more
advanced algorithms and techniques to improve detection accuracy and
efficiency.
• Real-time processing: Enhancing the system to process transactions in
real-time, reducing the time between detection and prevention.
• Cloud-based deployment: Scaling the system to handle large volumes of
data and transactions by deploying it on cloud infrastructure.
• Multi-industry applications: Adapting the system for use in various
industries, such as healthcare, insurance, and e-commerce.
• Integration with IoT devices: Incorporating data from IoT devices to
detect and prevent fraud in emerging technologies like smart homes and
cities.
conclusion
Payment card fraud is a massive problem for the
Banking sector. Hence, an effective fraud detection
system for card payments is needed by any bank or
financial institution to reduce the damages caused by
fraudulent activities.
In this research, we assumed that deviation from the
normal behavior of the cardholder could serve as the
basis for fraud detection. Our experiments showed that
the calculation of the similarity between existing
transactions in a cardholder's profile and test
transactions.
References
• Principal Component Analysis, Wikipedia Page,
https://en.wikipedia.org/wiki/Principal_component_analysis
• RandomForrestClassifier,
http://scikitlearn.org/stable/modules/generated/sklearn.ensemble.Ra
ndomForestClassifier. html
• ROC-AUC characteristic,
https://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area
_under_the_curve
• AdaBoostClassifier,
http://scikitlearn.org/stable/modules/generated/sklearn.ensemble.A
daBoostClassifier.html
• CatBoostClassifier,
https://tech.yandex.com/catboost/doc/dg/concepts/pythonreference
_catboostclassifier- docpage/
THANK YOU
