FRAUD DETECTION IN

INTERNET BANKING USING

MACHINE LEARNING
PRESENTED BY
K. SriNiketh Raj – 21WJ1A6937
K. Vineeth Reddy – 21WJ1A6936
Mohd. Mudaseer Mazharuddin –
21WJ1A6947
Under the Esteemed Guidance of
Ms. Snober Iqbal(Assistant professor)

Guru Nanak Institutions Technical Campus

(Autonomous)
OVERVIEW:
- ABSTACT
- INTRODUCTION
- EXISTING SYSTEM
- PROPOSED SYSTEM
- ALGORITHM USED
- SYSTEM ARCHITECTURE
- SYSTEM REQUIREMENTS
- CONCLUSION
ABSTRACT:
• Fraud detection in internet banking systems is crucial for maintaining security and
trust in financial institutions. Traditional rule-based approaches, while effective,
often fail to detect novel and evolving fraud patterns.
• This project proposes a hybrid fraud detection system that integrates supervised
machine learning models with unsupervised anomaly detection techniques, aimed at
identifying both known and unknown fraudulent activities. The model leverages
features extracted from transaction logs, including behavioral patterns, geographic
data, and temporal trends, to enhance fraud detection accuracy.
• Additionally, the system incorporates a "prudent learning" mechanism, enabling real-
time adaptation to emerging fraud types through online learning and novelty
detection. Preliminary experiments with real-world banking data show promising
results, highlighting the system’s ability to rapidly identify and respond to both
existing and unforeseen fraud scenarios. This approach represents a significant
advancement over brittle rule-based systems, offering a more flexible and dynamic
solution for modern fraud detection in internet banking.
INTRODUCTIO
N:
• Fraud detection in internet banking is a critical aspect of financial security, as online
transactions continue to grow at an exponential rate. Traditional fraud detection
systems often rely on rigid rule-based mechanisms, which, while useful for known
fraud patterns, struggle to identify new and evolving threats. These systems tend to
be brittle, with limited capability to adapt to emerging fraud behaviors that lie
outside their predefined knowledge.

• This project addresses these limitations by developing a hybrid fraud detection

system that combines rule-based logic with machine learning models. The system
integrates supervised learning techniques for identifying known fraud patterns with
unsupervised anomaly detection methods to catch novel, previously unseen fraud.
Additionally, a "prudent learning" framework is incorporated, allowing the model to
continuously adapt and learn from new data in real-time. By leveraging advanced
machine learning algorithms and anomaly detection mechanisms, this project aims
to create a more dynamic, resilient, and responsive solution for fraud detection in
internet banking.
EXISTING
SYSTEM:
• Traditionally, banks do not publishes specific details of their Fraud Detection (FD)
systems . Despite this fact, there have recently been a number of software vendors
who published some information about their Internet banking FD tools and the
market share of such tools in commercial banks worldwide. For example, the
Proactive Risk Manager (PRM) was reported to be used by the top 20 banks in the
world and in more than 40 countries

Support Vector Machine

(SVM):
• Support Vector Machine (SVM) is a powerful supervised learning algorithm used for
classification and regression tasks. It works by finding the optimal hyper plane that
best separates different classes in the feature space. The hyper plane is determined
by maximizing the margin, which is the distance between the hyper plane and the
closest data points from each class, known as support vectors.
Support Vector Machine (SVM):
• One of the key strengths of SVM is its ability to handle high-dimensional data
effectively, making it suitable for tasks with a large number of features. Additionally,
SVM can handle non-linear decision boundaries through the use of kernel functions,
which map the original feature space into a higher-dimensional space where classes
are more easily separable. Common kernel functions include linear, polynomial,
radial basis function (RBF), and sigmoid.
DRAWBACKS:
• Results shows less accuracy compared to random forest classifier.
• Less effective in predicting the outcome of a continuous variable.
PROPOSED SYSTEM:
• The proposed system is designed to enhance the detection of fraudulent activities
within banking transactions through the utilization of machine learning algorithms.
It begins by collecting transaction data, encompassing essential details such as
amounts, timestamps, merchant information, and customer particulars.
• Training the system involves the division of the pre-processed data into training and
testing sets, where machine learning algorithms like Logistic Regression, Random
Forest, and Decision Tree are trained to discern patterns indicative of fraudulent
behaviour.
ALGORITHM USED:
Random Forest
• Random Forest is a versatile and powerful ensemble learning algorithm that
operates by constructing a multitude of decision trees during training.
• Each tree in the forest independently learns to classify instances based on a subset
of features, and the final classification is determined by aggregating the predictions
of all trees. K-Nearest Neighbours (KNN) is a simple yet effective algorithm used for
both classification and regression tasks.
• It operates on the principle of similarity, where the classification of a data point is
determined by the class labels of its nearest neighbours in the feature space. KNN
does not require explicit training as it stores all training data points and their
corresponding labels, making it straightforward to implement and interpret.
SYSTEM ARCHITECTURE:
SYSTEM
REQUIREMENTS:
HARDWARE REQUIREMENTS

 PROCESSOR : Pentium i3 Processor

 RAM : 2GB DD RAM

 HARD DISK : 250 GB

SOFTWARE REQUIREMENTS

 BACK END : PYTHON

 OPERATING SYSTEM : WINDOWS 7

 IDE : Spyder3
CONCLUSION:
• This project has successfully developed a comprehensive fraud detection system for
internet banking by combining rule-based methods with advanced machine learning
techniques. Through the integration of Random Forest classifiers and Neural Networks,
the system effectively identifies both known fraud patterns and previously unseen
fraudulent behaviors.
• The pre-processing and feature selection stages optimized the data, ensuring that
relevant transaction features were used for accurate prediction. By using supervised
models for known fraud detection and anomaly detection methods for novel fraud
patterns, the system addresses both static and evolving threats. The hybrid approach
enhances detection accuracy, while techniques like anomaly detection ensure
adaptability in real-time environments.
• Performance analysis, including precision, recall, and other metrics, showed the model’s
effectiveness in handling imbalanced datasets, which is a key challenge in fraud
detection. This adaptive system continuously improves with new data, reducing false
positives and false negatives, while ensuring a robust defense against emerging fraud
tactics.
• Overall, this project presents a scalable, adaptive, and reliable fraud detection
framework that significantly enhances the security of online banking systems, providing
an innovative alternative to traditional static methods. It enables real-time fraud
THANK YOU……