2025 International Conference on Recent Advances in Electrical, Electronics, Ubiquitous Communication, and Computational Intelligence (RAEEUCCI) | 979-8-3503-9266-1/25/$31.

00 ©2025 IEEE | DOI: 10.1109/RAEEUCCI63961.2025.11048184

Credit Card Fraud Detection in Banking

Using Machine Learning
PULIMI MAHESH SURESH CHINNATHAMPY M PARTHIBAN M
Department of Electronics and Department of Electronics and Department of Electronics and
Communication Engineering, Communication Engineering, Communication Engineering, Vel
Vel Tech High Tech Vel Tech High Tech Tech High Tech Dr. Rangarajan Dr.
Dr. Rangarajan Dr. Sakunthala Dr. Rangarajan Dr. Sakunthala Sakunthala Engineering College,
Engineering College, Avadi, Chennai Engineering College, Avadi, Chennai - Avadi, Chennai -600062, India.
-600062, India. 600062, India. parthiban.ece@velhightech.co
pulimimahesh@velhightech.com sureshchinnathampy@velhightech.c om m

ARAVINTH R KARTHICK G
JEEVAN KUMAR C Department of Electronics and Department of Electronics and
Department of Electronics and Communication Engineering, Communication Engineering, Vel
Communication Engineering, Vel Tech High Tech Tech High Tech
Vel Tech High Tech Dr. Rangarajan Dr.Sakunthala Dr. Rangarajan Dr. Sakunthala
Dr. Rangarajan Dr. Sakunthala Engineering College, Avadi, Chennai - Engineering College, Avadi,
Engineering College, Avadi, Chennai 600062, India. Chennai -600062, India.
-600062, India.
jeevankumarc_ece21@velhightech.c om aravinthr_ece21@velhightech.com karthickg_ece21@velhightech.com

Abstract:
sectors. Yet, this convenience came at the price of
heightened credit card fraud, which proved to be a
Credit card fraud has become a main issue in banking, formidable challenge for banks and consumers as well.
causing big losses and diminishing trust among With every innovation by fraudsters in their strategies,
customers. Conventional methods of detection fail to
it became tough for the conventional fraud prevention
catch up with new tactics used by fraudsters. The use of
machine learning strategies in credit card fraud system to remain current. Accordingly, a more
detection has been discussed here. We apply complex sophisticated and responsive set of solutions to identify
algorithms such as logistic regression, decision trees, and capture fraud effectively is needed in an urgent
random forests, and neural networks to analyze manner. This paper discusses the use of machine
transaction data in order to identify patterns learning (ML) methods to solve the increasing issue of
characteristic of fraud patterns. The study focuses on credit card fraud within the banking sector [1].
the feature engineering, model estimation, and real-
time detection mechanisms to achieve improved Supervised learning [2], unsupervised learning, and
accuracy and reduce false positives. Experimental deep learning algorithms are quite helpful for fraud
results demonstrate that machine learning-based transaction identification. Supervised learning
techniques surpass traditional methods and offer an algorithms such as logistic regression, decision trees,
excellent solution to avoid fraud risk. This research and random forests can be trained on labeled
emphasizes the potential of machine learning to information to mark the transactions as fraud or real.
revolutionize fraud detection in banking for enhanced
Unsupervised learning algorithms such as anomaly
security and customer satisfaction.
detection and clustering are most suited to detect
Keyword: Credit Card Fraud, Machine Learning, unknown fraud patterns. Deep learning techniques [3],
Fraud Prevention, Financial Security, Anomaly such as neural networks, are most appropriate for
Detection, Imbalanced Data Classification. highly dimensional and complicated data and therefore
they function optimally in real-time fraud detection.
I. INTRODUCTION Deep learning techniques being used together can
The emergence of electronic transactions at a high enable banks to develop efficient systems that learn to
growth rate and extensive use of credit cards have detect changing patterns of fraud and minimize
completely transformed the financial and banking financial loss.

Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.
The use of machine learning to identify credit card Supervised learning algorithms have been widely used
fraud is accompanied by a number of challenges. One for credit card fraud detection due to the ability to label
of the significant problems is the class imbalance, the transactions as fraudulent or genuine based on
where the fraudulent transactions constitute a very labeled datasets [4]. It has been demonstrated through
small percentage of all transactions. This can cause research that logistic regression, SVM, and decision
biased models towards the majority class, leading to trees are effective algorithms for detecting fraud. For
ineffective fraud detection. Oversampling, under instance, ensemble algorithms like random forests and
sampling, and creating synthetic data (e.g., SMOTE) gradient boosting have been shown to be very accurate
are utilized to avoid this problem. For avoiding this in aggregating strengths of multiple models. However,
issue, oversampling, under sampling, and synthetic one of the largest supervised learning challenges is class
data generation (e.g., SMOTE) are applied. Feature imbalance in fraud datasets, where fraudulent
engineering is utilized to select and manipulate transactions significantly outnumber legitimate
feature attributes from transactional data for optimal transactions. To address this, techniques such as
model performance. Since any delay in detection can oversampling, under sampling, and synthetic data
lead to massive financial loss, real- time detection is creation (e.g., SMOTE) have been employed to balance
required. Through data analysis of transactions and the dataset as well as to improve model performance
testing several ML algorithms, we aim to develop a
low- false-positive, effective, and accurate real-time Unsupervised learning techniques have also been
model. The outcome will help banks use ML-based identified as having the ability to detect unknown
solutions to combat fraud. Lastly, machine learning in patterns of fraud without relying on labeled data.
fraud detection will enhance financial security and Clustering methods such as k-means and DBSCAN
boost the trust of customers in the banking sector. have been used to cluster similar transactions and
separate out outliers that could be indicative of fraud.
Algorithms such as isolation forests and auto encoders
have proven to detect well rare and unusual
transactions which do not represent usual behavior.
These unsupervised algorithms are particularly
beneficial for discovering new types of fraud that are
not based on well-known patterns and therefore
complement supervised learning algorithms.
Deep learning, which is a machine learning method [5],
Fig. 1. Credit Card Fraud Detection Using Machine Learning. has been found to be a highly effective approach in
detecting credit card fraud due to its ability to handle
II. RELATED WORKS
high-dimensional and complex data. Some of the neural
Credit card fraud detection has been a problem networks utilized for handling sequences of
explored at very long lengths in the last two years transactions and extracting important features to detect
with numerous techniques being employed to fraud are recurrent neural networks (RNNs) and
increase the efficiency and effectiveness of fraud convolutional neural networks (CNNs). Research has
detection mechanisms. Statistics and rule-based established that deep learning models are capable of
mechanisms were among early solutions that were state-of-the-art performance on fraud detection
most dependent on pre-defined thresholds and problems, particularly when augmented with methods
patterns in identifying suspicious transactions. such as transfer learning and attention mechanisms. Yet,
Whereas these techniques had helped in the the computational complexity and resource needs of
detection of well-known fraud patterns, they were deep learning models pose a problem for There have
unable to match the latest and refined means of also been recent experiments on integrating real-time
operating by fraudsters. Therefore, emphasis was fraud detection systems into banking infrastructure.
laid on newer techniques like machine learning, Stream processing platforms such as Apache Kafka and
which can process enormous amounts of transaction Apache Flink were used for real-time processing of
data and detect sophisticated, non-linear patterns transaction data and deployment of machine learning
characteristic of fraud. models in favor of real-time fraud detection.

Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.
 digital transactions increases, fraudsters constantly
come up with advanced methods to circumvent
conventional security features [6]. Traditional rule-
based fraud detection systems have difficulty keeping
up with changing fraudulent patterns and tend to have
high false-positive rates, where genuine transactions are
reported as fraud, or false negatives, where fraudulent
transactions are not detected. Moreover, the extremely
imbalanced nature of fraud detection datasets, where
fraudulent transactions constitute a minority of all
transactions, is a challenge for machine learning
models. Effectively detecting fraudulent transactions
without excessive false alarms is paramount to
preserving customer confidence and financial integrity.
Fig. 2. Data Structure Flow
Thus, there is a requirement for sophisticated machine
learning methods that can identify fraud in real-time,
Credit card fraud detection has been a well-studied learn dynamic fraud behavior, and enhance the overall
topic, with a variety of machine learning techniques accuracy and efficiency of fraud detection in banking
[7] applied to enhance security and reduce financial systems.
loss. Traditionally, rule-based fraud detection
systems have been widely used; these, however, are III. PROPOSED METHOD AND ALGORITHM
typified by high false positives and insensitivity to
evolving patterns of fraud. In a bid to overcome
these limitations, scholars have explored supervised This study proposes a machine learning-driven fraud
and unsupervised learning approaches. Supervised detection system that integrates supervised and
learning models such as Logistic Regression, unsupervised learning approaches to identify fraudulent
Decision Trees, Random Forest, Support Vector transactions with high accuracy. The proposed model
Machines (SVM), and Deep Learning models such will employ a mix of feature engineering, anomaly
as Convolutional Neural Networks (CNN) and detection, and ensemble learning to increase fraud
Recurrent Neural Networks (RNN) have reported detection effectiveness. Raw transaction data will be
encouraging outcomes in identifying fraudulent preprocessed for the first time by handling missing
transactions [8],[9]. Supervised learning models are values, encoding categorical variables, and solving data
trained on past transaction data with fraud labels, imbalance through techniques like Synthetic Minority
thereby enabling them to mark new transactions Over- sampling Technique (SMOTE)
efficiently [10]. The framework will employ supervised learning
methods like Decision Trees, Random Forest, Support
Recent studies have aggressively pursued real-time Vector Machines (SVM), and Deep Learning models
fraud detection via online learning models that like Long Short-Term Memory (LSTM) and
update upon identifying evolving patterns in fraud Convolutional Neural Networks (CNN) to identify
[11]. Graph- based methods have also been transactions as fraud or real. Besides this, other
investigated for detecting fraud networks based on unsupervised machine learning techniques like isolation
relationships between entities and transactions. forests and auto encoders will be used in order to
Federated learning has also been proposed as a identify any anomaly in the pattern of transactions. For
privacy- protecting alternative to allow different improving the overall detection rate and reducing false
banks to cooperatively detect fraud without positives, robustness of the model will be obtained by
exposing sensitive customer information [12]. using an ensemble approach with a list of classifiers.
Although machine learning greatly enhanced the A. Algorithm
performance of fraud detection, there remain issues
like dealing with imbalanced datasets, adversarial Bank credit card fraud detection is based on a set of
attacks, and dynamic fraud tactics. Upcoming machine learning algorithms to identify fraud
research focuses on explainable AI (XAI) to transactions in the right way. Supervised machine
promote interpretability and fairness in fraud learning algorithms like Logistic Regression, Decision
detection models. Tree, Random Forest, Support Vector Machines
(SVM), and Gradient Boosting
A. Problem Statements
Credit card fraud presents a significant challenge in
the banking industry, causing huge financial losses
to both banks and customers. As the number of

Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.
(XGBoost, LightGBM, CatBoost) are the most Data Set: Downloading datasets from Kaggle can be
popular algorithms for fraud classification. They are beneficial for data analysis, machine learning, and
trained on labeled samples of known fraud historical research. This dataset has 4,850 records and 11 fields,
transactions, and therefore they can detect true as with a size of around 319KB. It seems to deal with
well as fraud transactions. credit card transactions, with one record per
transaction. The fields have some information about
the transaction and cardholder. The "Unnamed: 0"
column likely is an index or auto-indexed ID of each
row. The "cc_num" column has the credit card number
or its masked form, and "category" has the category of
the transaction, e.g., grocery shopping, gas, online
shopping.The "amt" column captures the amount spent
on each transaction, and the "gender" column captures
the gender of the cardholder. The "is_fraud" column is
a binary flag, with 1 representing a fraudulent
transaction and 0 representing a valid one. The "age"
column contains the age of the cardholder, and the
"trans_month" and "trans_year" columns detail the
date of the transaction. Lastly, the "lat_dis" and
"long_dis" columns represent the geographic
distance (in latitude and longitude) between the
transaction location and the cardholder's known
Fig. 3. K-nearest neighbor classification location, which may aid in spotting suspicious activity
or fraud due to location irregularities.
Feature Selection and Engineering: This module
emphasizes identifying and selecting essential
transaction attributes for fraud detection. Significant
attributes including transaction amount, location,
timing, device information, and behavioral patterns of a
user are used to improve the accuracy of a model.
Reduction of dimension and improvement in computing
efficiency are attained through techniques like Principal
Component Analysis (PCA) and Recursive Feature
Elimination (RFE)
Model Training and Classification: Various machine
learning algorithms, including Decision Trees, Random
Forest, Support Vector Machines (SVM) are
implemented in this module. Supervised learning is
used for labeled transaction data, while unsupervised
Fig. 4. Decision Tree techniques identify anomalies without predefined fraud
labels.
B. Module List and Descriptions Anomaly Detection and Fraud Identification: This
Data Collection and Preprocessing: This module module utilizes unsupervised learning algorithms like
is for gathering credit card transaction data samples Isolation Forest, One-Class SVM, and Clustering
from banking statements and open sources. algorithms (K-Means, DBSCAN) to identify abnormal
Preprocessing data operations such as handling transaction patterns that could suggest fraudulent
missing values, removing duplicates, transforming activity. These models assist in detecting new fraud
categorical variables into numerical variables, and patterns which could escape supervised models.
normalizing numeric features are employed. Also,
methods such as Synthetic Minority Over-sampling
Technique (SMOTE) are employed to solve the
imbalance in class and provide an efficient training
set.

Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.
Real-time Fraud Detection and Alert System: The
framework combines stream data processing
platforms to identify fraud in real-time. In case of a
transaction suspected of being fraudulent, an alert is
triggered and subsequent verification processes, like
multi-factor authentication, are invoked to block
unauthorized transactions.

Performance Evaluation and Optimization: The

last module assesses model performance on
measures like accuracy, precision, recall, F1-score,
and ROC- AUC curves. Hyper parameter tuning
methods like Grid Search and Random Search are
employed to optimize the model and enhance fraud
Detection.
IV. RESULTS
The machine learning approach proposed for Fig. 5. Accuracy Graph
detecting credit card fraud in banking is made up of Anomaly Detection & Fraud Identification: The
several stages for the purpose of achieving accurate trained models categorize transactions into fraudulent
and efficient. or legitimate using learned patterns. The blended
Data Acquisition & Preprocessing: The procedure techniques that use both supervised and unsupervised
begins with the collection of transactional data from techniques help accuracy and fraud discovery
bank statements or publicly available datasets. The improvement. Other ensemble learning approaches like
data is preprocessed by handling missing values, Boosting and Stacking help increase the performance of
removing duplicates, and encoding categorical detection.
features. Since fraud detection datasets are usually Real-Time Detection & Alert Generation: A real-
highly imbalanced, techniques like SMOTE and time fraud detection system is implemented in a
under sampling are used to balance the dataset so banking system, where streaming data is scanned in real
that the model does not lean towards non- time for dubious fraudulent transactions. In case a
fraudulent transactions. doubtful transaction is detected, the system initiates
automatic alerts and security measures like OTP
Feature Selection & Engineering: Relevant verification or locking of an account to avert fraud.
transaction attributes, including transaction value,
frequency, device usage, geographical location, and Model Evaluation & Performance Optimization: The
time-based patterns, are derived. Feature scaling and value of fraudulent detection models is measured by
transformation methods, like Min-Max scaling or Precision, Recall, F1-score, Confusion Matrix, and
Standardization, are used to maintain consistency in AUC-ROC curve. Methods of hyper parameter tuning
data distribution. such as Grid Search, Bayesian Optimization, and
Neural Architecture Search (NAS) are applied for
Model Selection & Training: The data set is improving model performance.
separated into training and testing sets, and machine
learning models are trained. It is classified through
supervised learning models such as Logistic TABLE I. ALGORITHM CLASSIFIER
Regression, Decision Trees, Random Forest, SVM.
Anomaly detection is performed through the
unsupervised learning algorithms such as Auto
encoders and Isolation Forest.

Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.
 [4] V. N. Dornadula and S. Geetha, “Credit Card Fraud Detection
using Machine Learning Algorithms,” Procedia Comput. Sci., vol.
165, pp. 631–641, 2019, doi: 10.1016/j.procs.2020.01.057.

[5] A. R. Khalid, N. Owoh, O. Uthmani, M. Ashawa, J. Osamor, and

J. Adejoh, “Enhancing Credit Card Fraud Detection: An Ensemble
Machine Learning Approach,” Big Data Cogn. Comput., vol. 8, no. 1,
p. 6, Jan. 2024, doi: 10.3390/bdcc8010006.

[6] R. Sailusha, V. Gnaneswar, R. Ramesh, and G. R. Rao, “Credit

Card Fraud Detection Using Machine Learning,” in 2020 4th
International Conference on Intelligent Computing and Control
Systems (ICICCS), IEEE, May 2020, pp. 1264–1270. doi:
10.1109/ICICCS48265.2020.9121114.

[7] Z. Zaffar, F. Sohrab, J. Kanniainen, and M. Gabbouj, “Credit Card

Fig. 6. ROU Curve
Fraud Detection with Subspace Learning-based One-Class
Classification,” in 2023 IEEE Symposium Series on Computational
V. CONCLUSION
Intelligence (SSCI), IEEE, Dec. 2023, pp. 407–412. doi:
Machine learning-based credit card fraud detection 10.1109/SSCI52147.2023.10372038.
in banking has been an effective tool for detecting [8] M. A. Talukder, R. Hossen, M. A. Uddin, M. N. Uddin, and U. K.
and preventing fraudulent transactions in real time. Acharjee, “Securing transactions: a hybrid dependable ensemble
Through the use of supervised, unsupervised, banks machine learning model using IHT-LR and grid search,”
are able to identify sophisticated fraud patterns that Cybersecurity, vol. 7, no. 1, p. 32, Nov. 2024, doi: 10.1186/s42400-
rule-based systems are unable to detect. The 024-00221-z.
application of techniques such as anomaly detection,
[9] D. Rzayeva and S. Malekzadeh, “A Combination of Deep Neural
ensemble learning, and feature engineering
Networks and K-Nearest Neighbors for Credit Card Fraud
enhances the accuracy and reliability of fraud
Detection,” May 2022, [Online]. Available:
detection systems. In addition, real-time processing http://arxiv.org/abs/2205.15300.
and automatic alerts allow financial institutions to
take prompt action against suspicious activities, [10] D. H. M. de Souza and C. J. Bordin, “Ensemble and Mixed
minimizing financial losses and boosting customer Learning Techniques for Credit Card Fraud Detection,” Dec. 2021,
trust. As the pace of innovative fraudulent [Online]. Available: http://arxiv.org/abs/2112.02627.
techniques accelerates, robust and elastic machine [11] F. K. Alarfaj, I. Malik, H. U. Khan, N. Almusallam, M. Ramzan,
learning algorithms will become even more crucial and M. Ahmed, “Credit Card Fraud Detection Using State-of-the-Art
in future years. Explainable AI, federated learning, Machine Learning and Deep Learning Algorithms,” IEEE Access,
blockchain protection, and quantum computing will vol. 10, pp. 39700–39715, 2022, doi:
steadily make fraud detection capabilities more user- 10.1109/ACCESS.2022.3166891.
friendly with assurances of privacy and
transparency. As banks continue the evolution and [12] S. Khatri, A. Arora, and A. P. Agrawal, “Supervised Machine
Learning Algorithms for Credit Card Fraud Detection: A
consolidation of new technology, they are able to
Comparison,” in 2020 10th International Conference on Cloud
develop an even more secure and efficient payment
Computing, Data Science & Engineering (Confluence), IEEE, Jan.
process, making fraud risk obsolete and further 2020, pp. 680–683. doi: 10.1109/Confluence47617.2020.9057851.
increasing overall transaction security.
ACKNOWLEDGMENTS
The authors acknowledge the HOD, Department of
Electronics and Communication Engineering, the
management of Vel Tech High Tech Dr. Rangarajan
Dr. Sakunthala Engineering college, Chennai, for their
continuous encouragement.

REFERENCES
[1] S. Patil, V. Nemade, and P. K. Soni, “Predictive Modelling
For Credit Card Fraud Detection Using Data Analytics,”
Procedia Comput. Sci., vol. 132, pp. 385–395, 2018, doi:
10.1016/j.procs.2018.05.199.

[2] E. Ileberi, Y. Sun, and Z. Wang, “A machine learning based

credit card fraud detection using the GA algorithm for feature
selection,” J. Big Data, vol. 9, no. 1, p. 24, Dec. 2022, doi:
10.1186/s40537-022-00573-8.

[3] I. D. Mienye and N. Jere, “Deep Learning for Credit Card

Fraud Detection: A Review of Algorithms, Challenges, and
Solutions,” IEEE Access, vol. 12, pp. 96893–96910, 2024, doi:
Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.
10.1109/ACCESS.2024.3426955.
Authorized licensed use limited to: VIT University. Downloaded on July 09,2025 at 16:39:17 UTC from IEEE Xplore. Restrictions apply.