SEISENSE Journal of Management

Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59

Research Article

A Review of Machine Learning Applications for

Credit Card Fraud Detection with A Case study
Zahra Faraji1 *

1Financial Service Analytics Institution, University of Delaware, United States

* Corresponding author: [email protected]

Purpose - This paper aims to highlight the widely used supervised

Article History
techniques applied for fraud detection. In addition, this paper aims
Received 2022-01-14
to apply some techniques to evaluate their performance on real-
Revised 2022-02-10
world data and develop an ensemble model as a potential solution
Accepted 2022-02-03
for this problem.
Published 2022-02-15
Design/Methodology- Different techniques applied in this study
Keywords
for fraud detection purposes are logistic regression, decision tree,
Machine learning,
random forest, KNN, and XGBoost. The confusion matrix gives
Credit card fraud detection,
information about the assignment of inputs to the different classes.
Artificial Neural Networks,
This study uses precision and recall to evaluate the performance,
Logistic regression,
calculated based on the confusion matrix.
Random Forests,
XGBoost Findings- XGBoost is the fastest and is expected to have the best
How to cite? performance; however, it is only outperforming the random forest
Faraji, Z. (2022). A Review of Machine in terms of accuracy, precision, recall, and f1-score. In general, the
Learning Applications for Credit Card KNN and logistic regression have better performance, which
Fraud Detection with A Case study. means they better detect fraudulent transactions.
SEISENSE Journal of Management, Practical Implications- The new model can be applied to new data
5(1), 49-59. instead of the previous techniques.
doi:https://doi.org/10.33215/sjom.v5i1.
770

Copyright © 2022 The Author(s)

49
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Introduction
Credit cardholders are encouraged to use their card over other ways of payment, such as debit cards. Indeed,
there are some advantages to using credit cards compared with other payment methods; for example, they are
so easy and fast to use, which support national or international transactions and withdrawing cash (Al Rubaie,
2021). However, some associated risks increase the attention for better credit card use management. Credit
cards are an easy target for fraud, but detecting fraudulent transactions is usually an easy task (Al Rubaie, 2021;
Dornadula, 2019). There are so many online businesses that only accept credit card transactions high usage of
credit cards for online transactions led to the high number of fraudulent transactions.
Fraud detection is crucial, and the increasing trend of credit card online transactions makes strong prevention
and detection techniques necessary. The total number of Visa and Master card users was about 2,183 million
worldwide in 2020. Over 30% of the Visa card users and 24% of Master card users were in the United States
(Alkhatib, 2021). The increasing number of cashless transactions resulted in an increasing trend of fraud rate
for Card Not Present (CNP) transactions (Murli, 2015). In addition, the COVID pandemic led more people
and merchants to use online transactions, highlighting the need to prevent and detect fraud. Consequently,
Researchers and financial institutions are continuously seeking more efficient techniques to increase the safety
of online transactions. Financial institutions are using prevention and detection mechanisms to check all the
transactions, and Machine Learning (ML) techniques have been so helpful for this purpose.
There are different ML techniques to tackle credit card fraud detection, but we can classify them into main
groups, including supervised, unsupervised, and reinforcement learning. The supervised learning techniques are
applicable for classification and prediction problems, and data should be labeled for these techniques. This
group contains techniques such as Support Vector Machine (SVM), Logistic Regression, Decision Tree, Naïve
Bayes, K-Nearest Neighbor, Random Forest, Artificial Immune System, and Artificial Neural Network. On the
other hand, the unsupervised learning techniques work with the unlabeled data and cluster the inputs based on
their similarities. Some unsupervised ML techniques are K-means, Hidden Markov Model, Genetic Algorithm,
Gradient Descendent, and DBSCAN (Zareapoor, 2015).
This paper is organized as follows: Section 2 gives information about the related credit card fraud detection
techniques. Section 3 describes the experimental setup approach by explaining some data challenges, required
pre-processing, and the different classifiers used to detect fraudulent credit card transactions. Section 4 presents
the experimental results and discussion, and Section 5 concludes this work and provides some suggestions for
future studies.

Literature Review
This section reviews some of the most important and widely used supervised classification techniques for credit
fraud detection. Credit card fraud is a binary problem, as a transaction can be normal or fraudulent. The goal
for the classification techniques is to classify a transaction based on the available features as fraud or legal.
Artificial Neural Network
Artificial Neural Network (ANN) imitates the human brain based on interconnected nodes or neurons. Each
ANN has input, hidden, and output layers. Using ANN, the computer can learn from data and humanly make
decisions. (Mehndiratta S. &., 2019). The neurons receive and process the signals and send them to the
subsequent neurons for further process. There is a weighted connection between the active nodes in the
associated layers. ANN has been widely used because of its ability to extract patterns in complex data. (Lim,
2021). Many researchers applied this algorithm for credit card fraud detection purposes (Murli, 2015; Paruchuri,

50
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

2017). The Convolution Neural Network (CNN) and Recurrent Neural Network (RNN) are two types of neural
networks implemented by the researcher to detect fraudulent activities. (Tiwari, 2021).
Genetic Algorithm (GA)
The genetic algorithm can optimize different problems such as fraud detection and prevention for online
transactions (Paruchuri, 2017). This method is a subset of the natural evolution idea to reproduce the fitter or
stronger individuals for the next generation (Tiwari, 2021).
Logistic Regression
The logistic regression model is a generalized linear technique used when the predictive variable is binary. This
technique uses the logistic function to model the probability for different classes (Tiwari, 2021). Logistic
regression has been widely used for fraud detection (Ngai, 2011). Some studies showed that logistic regression
could outperform other credit card fraud data (Paruchuri, 2017).
Decision Tree
A decision tree is a predictive model for classification and regression problems that map the inputs to the
possible classes. This supervised technique has a tree-like structure that contains root nodes, other nodes that
split the data based on the features, and leaves. At each node, the classes in the dataset are separated based on
some conditions on the features. The splitting process achieves the fullest purity (Paruchuri, 2017). This
technique is widely used for credit card fraud detections (Ngai, 2011; Paruchuri, 2017).
Random Forest
Random forest is another supervised technique that uses the bagging idea to improve the results by combining
multiple single trees. This approach uses a random subset of each tree's features and a training dataset to
overcome the disadvantages of a single decision tree (Tiwari, 2021). This method has been used for credit card
fraud detection for online and offline transactions (Paruchuri, 2017).
Support Vector Machines (SVM)
SVM is another supervised method that is applicable for classification and regression problems. This technique
can handle high-dimensional data by transforming a non-linear input into a linear task. SVM has been applied
to detect different types of fraud in banking industries (Paruchuri, 2017). SVM creates the decision boundary
to segregate the classes.
K-Nearest Neighbors (KNN)
KNN is a supervised method used for classification and regression challenges. This method classifies each input
based on the nearest feature space in the training dataset without generalization (Lim, 2021; Mehndiratta S. &.,
2019). This technique is fast but, as mentioned, doesn’t make any generalizations. Indeed, if the K-nearest
neighbors are fraudulent transactions, it is highly possible that a new instance is classified as fraudulent.
Naïve Bayesian Classifier
This method is only applicable for classification problems that leverage the Bayesian rules to calculate the
conditional probabilities of the classes (Tiwari, 2021). The class with the highest conditional probabilities will
be assigned to the instances. An important assumption for this model is that the effect of a feature on a given
label is unrelated or independent from the value of the other features (Mehndiratta S. &., 2019).
Gradient Boosted Trees:
This is a supervised technique for classification and regression problems using an ensemble of decision trees.
It uses the gradient boosting framework and builds sequences of weak models or trees, and at each iteration, it

51
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

tries to predict the error of the previous model (Mishra, 2018). XGBoost is based on gradient boosted decision
trees, which are also applicable for categorical and continuous target variables. It is more regularized than
gradient boosted trees for more generalization, resulting in better speed and performance (Meng, 2020).
Hidden Markov Model (HMM)
This stochastic model contains unobserved or hidden states that can be predicted by the underlying Markov
Chain. Indeed, the only available information is the observational data, and there is no information related to
the states. This model's main assumption is that future event are only related to the current state but not to the
previous states (Tiwari, 2021).
Another way to tackle the fraud detection problem is combining or using multiple machine learning techniques
to perform better. Different studies compared multiple methods together and even made a hybrid technique to
boost the model (Faraji Z. , 2020; Faraji Z. , 2020; Asgharizadeh, 2014; Faraji, Z., Fleischhacker, A., 2020).
Research tackled the credit card fraud detection problem with three supervised techniques, support vector
machines, logistic regression, and random forests. One study compared the naïve Bayes model and logistic
regression for fraud detection. The findings revealed that the naive Bayes model could converge faster, but the
discriminative logistic regression has the lower asymptotic error (Ng, 2002). Another research studied the three
models: logistic regression, decision tree, and the neural network for fraud detection. The findings showed the
decision tree underperformed the other models (Shen, 2077). However, in another study, the decision tree
outperformed the SVM and had better performance detecting fraudulent instances (Sahin, 2011)e findings
suggested that the logistic regression had better performance. Still, SVM considered more fraudulent instances
in the training data (Bhattacharyya, 2011). Another study applied 10 ML techniques and compared their
performance using the accuracy and confusion matrix. They applied Decision Tree, KNN, SVM, Logistic
Regression, Random Forest, Naïve Bayes, Gradient Boosting, XGB Classifier, and Stacking Classifier to detect
the fraudulent instances logistic regression showed better performance (Dhankhad, 2019).
Different standard and hybrid machine learning techniques were used to detect credit card fraud activities in
one study. The hybrid model applied the AdaBoost with majority voting technique for a real-world dataset in
which noises were added to test the model robustness. Random Forest showed the best performance compared
to other models in this study (Randhawa, 2018). Based on the literature, different machine learning techniques
can be applied to detect fraud activities for the credit card. Each technique has its benefits and disadvantages
(Faraji, Z., Fleischhacker, A., 2020; Faraji Z. , in press); however, a researcher should choose a technique based
on the available data and associated feasibilities. For example, SVM requires lots of memory and considerable
time to train the data. All the attributes should be independent for naïve Bayes; KNN needs scaled data, which
is not applicable for high-dimensional data (Uchhana, 2021). Indeed, it is necessary to consider the data and
computation limitations to choose the appropriate technique (Faraji, Z., Fleischhacker, A., 2020; Faraji Z. ,
2020).

Research Method
Data Challenges
Usually, the dataset for fraud detection is unbalanced as the number of fraudulent transactions is considerably
less than the normal transactions. All the previously mentioned methods can predict the majority class in the
data, but the fraudulent class is considered noises. Therefore, misclassification for minority class instances is
high while using the standard techniques for fraud data (Mishra, 2018). However. The goal is to classify the
minority class with good performance. One solution is to apply the Synthetic Minority Oversampling Technique
(SMOTE) for raw data before the modeling. SMOT oversamples the minority class and under samples the

52
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

majority class. Indeed, the new dataset is balanced. This method creates "synthetic" from the minority class
using the KNN technique (Mishra, 2018).
Even if the model classifies all the instances as normal transactions with highly imbalanced fraud data, the
accuracy would be high but misleading. Therefore, accuracy is a weak performance metric for fraud data. This
study uses precision and recall to evaluate the performance, calculated based on the confusion matrix. The
confusion matrix gives information about the assignment of inputs to the different classes. The following table
shows the confusion matrix (Al Rubaie, 2021).
Table 1. Confusion Matrix
Confusion Matrix
Positive (Fraud) Negative (Normal)
Positive (Fraud) True Positive (TP) False Negative (FN)
Negative (Normal) False Positive (FP) True Negative (TN)

The performance metrics can be calculated based on the confusion matrix as follows:
True Positive (TP) = number of fraud transactions predicted as fraud
True Negative (TN) = number of legal transactions predicted as legal
False Positive (FP) = number of legal transactions predicted as fraud
False Negatives (FN) = number of fraud transactions predicted as legal
These values are applied to compute the following (Zareapoor, 2015):

TP + TN
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =
TP + FP + TN + FN

TP
𝑅𝑒𝑐𝑎𝑙𝑙 =
TP + FN

TP
𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛 =
TP + FP

TP
𝐹1 =
1
TP + (FP + FN)
2

Data Description
The data for this study is public data available on the Kaggle website. The data contains 284,807 transactions in
a two-day duration that only a tiny fraction of that is labeled as fraudulent transactions. There are 28 features
in the dataset transformed by principal component analysis (PCA). The transaction amount is another feature
that needs to be scaled before training the model. The next table summarizes the transaction distribution, and
Figure 1 visualizes the imbalance data.
Table 2. Data Description
Total Transaction Legal Fraud Fraud rate
284, 807 284315 492 0.173%

53
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Figure 1. Fraud Data Distribution

Our Approach
The different classification techniques are applied in this study for fraud detection purposes are logistic
regression, decision tree, random forest, KNN, and XGBoost. Their performances are compared to see which
model can better extract the relationship between the features and detect fraudulent transactions. The logistic
regression estimates the probabilities by using a logistic regression equation to model the relationship between
the dependent and independent variables. The decision tree is another classification model that is applied to
the data. This model has some advantages, such as being easy to interpret the results and visualize the tree,
making it easier to communicate with non-technical persons. The random forest is applied to the dataset as an
ensemble model that includes more trees. The final results are based on the most frequent predictions by
individual trees. The gradient boosting method is a tree-based learning algorithm with a sequence of decision
trees built on the previous trees' errors. This model is fast and can be an excellent choice for fraud data. In
addition, cross-validation and hyperparameters tuning is applied for all the models in this study to find the best
fit and avoid overfitting.
After training all the classifiers, a new ensemble model will be applied as a voting classifier to combine all the
other classification techniques. The objective is to reduce the errors of single models, which helps the ensemble
model make better predictions compared with the individual classifiers. If all the classifiers are considered as
𝐶1 , 𝐶2 , 𝐶3 , 𝐶4 , 𝑎𝑛𝑑 𝐶5 , then the final classifier will take the votes as the majority of votes as the final prediction
or 𝐶𝑡 .
𝐶𝑡 = 𝑀𝑎𝑗𝑜𝑟𝑖𝑡𝑦{𝐶1 , 𝐶2 , 𝐶3 , 𝐶4 , 𝐶5 }

Results
The next table shows the model's performance before balancing the data. According to Table 3, the uneven
instances in the data resulted in poor performance for most of the models. Figure 2 shows the performance of
the XGBoost model, which is overperforming the other models. Most accuracy scores are high as most normal
transactions are classified correctly in imbalanced data. However, by looking at the other metrics, it is clear that
the precision, recall, and f1-score are lower than the accuracy. Also, the XGBoost model is overperforming the
other techniques based on all the metrics.

54
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Figure 2. Confusion Matrix for Unbalanced Data (XGBoost Model)

Table 3. Mode Performance for Unbalanced Data
Methods Accuracy Precision Recall F1-score
Logistic regression 0.81 0.90 0.63 0.74
Decision tree 0.99 0.77 0.76 0.77
Random forest 0.99 0.94 0.78 0.85
KNN 0.86 0.91 0.72 0.81
XGBoost 0.99 0.92 0.82 0.87
In the next step, the data is balanced using the SMOT method to oversample the fraud transactions and under-
sample the normal transactions. Figure 3 shows the new distribution for the dataset.

Figure 3. Fraud Data Distribution for Balanced Data

After balancing the data, all the previous machine learning techniques are applied. The next figure shows the
confusion matrix for the XGBoost model, which illustrates the performance improvement after leveraging the
SMOT method.

55
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Figure 4. Confusion Matrix for Balanced Data (XGBoost Model)

Table 4 gives detailed information about the performance measurements for all the applied models. A new
ensemble model is also applied as a voting classifier to combine all the other classification techniques. The idea
is that the ensemble model is stronger than the single model. The results are based on voting on the predicted
classes, aiming to reduce the error. According to Table 4, the voting classifier model outperforms the other
models.
Table 4. Mode Performance for Balanced Data
Methods Accuracy Precision Recall F1-score
Logistic regression 0.98 0.98 0.93 0.96
Decision tree 0.99 0.77 0.76 0.77
Random forest 0.99 0.93 0.78 0.85
XGBoost 0.99 0.93 0.83 0.88
KNN 0.98 0.98 0.94 0.96
Ensemble 0.99 0.99 0.99 0.99

The logistic regression performance is better than the decision tree, and the XGBoost outperforms the random
forest in terms of accuracy, precision, recall, and f1-score. The KNN and logistic regression have better
performance than other models. Their low false-negative rates mean better at capturing the fraudulent
transactions and extracting the data pattern. The ensemble model gains better results by reducing the errors of
the other models. Figure 5 visualizes the performance measurements.

Model Performnace
1.5

0.5

0
Logistic Decision tree Random forest XGBoost KNN Ensemble
regression

Accuracy Precision Recall f1-score

Figure 5. Model Performance

56
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Discussion
The ensemble model gains better results by reducing the errors of the other models. The logistic regression is
the most straightforward technique in this paper, but its performance is better than the decision tree. The
XGBoost is the fastest and expected to have the best performance; however, it is only outperforming the
random forest in terms of accuracy, precision, recall, and f1-score. In general, the KNN and logistic regression
have better performance, which means they better detect fraudulent transactions. These models are more
straightforward than XGBoost, highlighting that the model complexity doesn’t guarantee good performance.
Each dataset has its features and requires investigation to find the appropriate model.

Conclusion
This paper reviewed the machine learning techniques for credit card fraud detection problems. Fraud is an
important problem for financial institutions; therefore, applying a model that can handle the data fast and
efficiently is critical. Balancing the data is so crucial to achieving a stable and generalized approach. Being
familiar with the different algorithms can be helpful to make a better decision to choose the right technique. In
this paper, five different supervised techniques were applied for public Kaggle data and their performance was
investigated with imbalance data and the balanced data. in addition, a new ensemble model was applied to
improve the performance of the individual classifiers.
Limitations and Future Research
There are some limitations to this study; one limitation is the data. The data for this study was limited to one
financial institution, which indicates that the results cannot be generalized for all the banks or financial
institutions. Future studies can investigate the ML techniques with larger datasets. Also, another limitation is
that the unsupervised techniques are not used in this study. In future studies, the new ensemble model can be
applied for a group of supervised and unsupervised techniques to see the possibility of performance
improvement. In future works, the unsupervised machine learning algorithms can be reviewed and compared
with supervised techniques. In addition, the data for this study has only numerical features. However, other
types of the data, such as textual data, can be beneficial to improve the fraud detection process.
Funding: This research received no external funding
Conflicts of Interest: The authors declare no conflict of interest

References
Al Rubaie, E. M. (2021). Improvement in credit card fraud detection using ensemble classification technique and user
data. International Journal of Nonlinear Analysis and Applications, 12(2), 1255-1265.
Alderson, A. S., Beckfield, J., & Nielsen, F. (2005). Exactly how has income inequality changed? Patterns of
distributional change in core societies. International Journal of Comparative Sociology, 46(5-6), 405-423.
Alkhatib, K. I.-A. (2021). Credit Card Fraud Detection Based on Deep Neural Network Approach. 12th International
Conference on Information and Communication Systems (ICICS) (pp. 153-156). IEEE.
Anjum, A., Yasmeen, K., & Khan, B. (2011). Performance Appraisal Systems In Public Sector Universities Of Pakistan.
International Journal of Human Resource Studies, 1(1), 41-51. Retrieved from
http://www.macrothink.org/journal/index.php/ijhrs
Armstrong M. (2006). A Handbook of Personnel Management Practice (6th ed.). London: Kogan Pag.
Asgharizadeh, E. S. (2014). Supply chain performance measurement by combining criterions of SCOR, Gunasekaran
and BSC models with REGIME technique. Journal of Basic Applied Scientific Research, 4(3), 309-320.
Bhattacharyya, S. J. (2011). Data mining for credit card fraud: A comparative study. Decision support systems, 50(3),
602-613.
Ciarniene, R. and Vienazindiene, M. (2010). Critical Issues for Compensation and Incentives Management: Theoretical
Approach. Management Theory and Studies for Rural Business Infrastructure Development, 5(24), 15-21.

57
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Danish, R.Q, & Usman, A. (2010). Impact of Reward and Recognition on Job Satisfaction and Motivation: An
Empirical Study from Pakistan.,. International Journal of Business and Management, 5(2), 159-167.
Dhankhad, S. M. (2019). Credit card fraud detection using machine learning algorithms. Procedia computer science,
165, 631-641.
Dornadula, V. N. (2019). Credit card fraud detection using machine learning algorithms. Procedia computer science,
165, 631-641.
Faraji, Z. (2020). The Causal Analysis of Financial Distress Risk and Performance. American International Journal of
Business Management, 3(5), 5.
Faraji, Z. (2020). Measuring Corporate Diversification Effect Sizes from Observational Data Using Bayesian Causal
Approaches. University of Delaware.
Faraji, Z. F. (2020). Overstatement of firm diversification effect on bankruptcy risk reduction. International Journal of
Applied Decision Sciences, 13(3), 267-285.
Faraji, Z. (in press). Investigate the causal effect of diversification strategy on risk-adjusted performance using Bayesian
additive regression trees. International Journal of Information and Decision Sciences.
Gazioglu, S. and A. Tansel. (2006). Job Satisfaction in Britain: Individual and Job Related Factors. Applied Economics,
8, 1163-1171.
Iqbal, M. Z., Arif, M. I., and Abbas, F. (2011). HRM Practices in Public and Private Universities of Pakistan: A
Comparative Study. International Education Studies, 4(4), 215-222.
Karatepe, O. M., & Vatankhah, S. (2015). High-performance work practices, career satisfaction, and service recovery
performance: a study of flight attendants. Tourism Review, 70(1), 56-71.
Khurram,S., & Sajid Bashir. (2008). Impact of HR Practices on Perceived Performance of. International Review of
Business Research Papers, 4(2), 302-315.
Kwenin, D.O., Muathe, S., & Nzulwa, R. (2013). The Influence of Employee Rewards, Human Resource Policies and
Job Satisfaction on the Retention of Employees in Vodafone Ghana Limited. European Journal of Business
and Management, 5(12), 2222-2839.
Kwenin, D.O., Muathe, S., & Nzulwa, R. (2013). The Influence of Employee Rewards, Human Resource Policies and
Job Satisfaction on the Retention of Employees in Vodafone Ghana Limited. European Journal of Business
and Management, 5(12), 2222- 2839.
Lai, H. ( 2012). A Study of the Relationship among the CompensationManagement, Operational and Organisational
Climate. International Journal of Organisatonal Innovation, 231-249.
Lamba, S., and Choudhary, N. (2013). Impact of HRM Practices on Organizational Commitment of Employees.
International Journal of Advancements in Research & Technology, 2(4), 407-423.
Lim, K. S. (2021). A Review of Machine Learning Algorithms for Fraud Detection in Credit Card Transaction.
International Journal of Computer Science & Network Security, 21(9), 31-40.
Majumder, T. H. (2012). Human Resource Management Practices and Employees’ Satisfaction Towards Private
Banking Sector in Bangladesh. International Review of Management and Marketing, 2(1), 52-58.
McIvor,R. (2007). Outsourcing and the spin-off arrangement: lessons from a utility company. Journal of General
Management,, 33(1), 51-70.
Mehndiratta, S. &. (2019). Credit card fraud detection techniques: a review. Int. J. Computer Sci. Mobile Computing,
8(8), 43-49.
Mehndiratta, S. &. (2019). Credit card fraud detection techniques: a review. Int. J. Computer Sci. Mobile Computing,
8(8), 43-49.
Mello, J.A. (2007). Strategic Human Resource Management (2nd ed.). India: Cengage.South-Western.
Memon, S.B., Panhwar, A.I. & Rohra, L.C. (2010). Investigating the Mediating Role of Human ResourPolicies in
Employee Retention. Australian Journal of Basic and Applied Science, 4(8), 4046-4057.
Memon, S.B., Panhwar, A.I. & Rohra, L.C. (2010). Investigating the Mediating Role of Human Resource Policies in
Employee Retention. Australian Journal of Basic and Applied Sciences, 4(8), 4046-4057.
Memon, S.B., Panhwar, A.I. & Rohra, L.C. (2010). Investigating the Mediating Role of Human Resource Policies in
Employee Retention. Australian Journal of Basic and Applied Sciences., 4(8), 4046-4057.
Meng, C. Z. (2020). A case study in credit fraud detection with SMOTE and XGboost. In Journal of Physics:
Conference Series (Vol. 1601, No. 5, p. 052016). IOP Publishing.

58
SEISENSE Journal of Management
Vol 5 No 1 (2022): DOI: https://doi.org/10.33215/sjom.v5i1.770 , 49-59
Research Article

Mishra, A. &. (2018). Credit card fraud detection on the skewed data using various classification and ensemble
techniques. In 2018 IEEE International Students' Conference on Electrical, Electronics and Computer Science
(SCEECS) (pp. 1-5). IEEE.
Muhammad Usman, W. A. (2015). Role of Board and Firm Performance in Determination of CEO Compensation:
Evidence from Islamic Republic of Pakistan. Pakistan Journal of Commerce and Social Sciences, 9(2), 641-657.
Murli, D. J. (2015). Credit card fraud detection using neural networks. International Journal of Students’ Research in
Technology & Management, 2(2), 84-88.
Ng, A. Y. (2002). On discriminative vs. generative classifiers: A comparison of logistic regression and naive bayes. In
Advances in neural information processing systems, 841-848.
Ngai, E. W. (2011). The application of data mining techniques in financial fraud detection: A classification framework
and an academic review of literature. Decision support systems, 50(3), 559-569.
Onukwube, H, N. (2012). Correlates of job satisfaction amongst quantity surveyors in consulting firms in Lagos,
Nigeria. . Australasian Journal of Construction Economics and Building, 12(2), 43-54.
Padala, S.R. (2011). Employees‟ Job Satisfactions and Organisational Commitment in Nagarjuna Fertilizers and
Chemicals Limited, India. ,Journal of Research in International Business and Management, 1(1), 17-27.
Paruchuri, H. (2017). Credit Card Fraud Detection using Machine Learning: A Systematic Literature Review. ABC
Journal of Advanced Research, 6(2), 113-120.
Plant, T. (2006). The Performance Measurement Paradox in Local Government Management in Pm. Public
Management, 88(4), 16-20.
Qasim, S. Cheema, F.E.A., & Syed, N.A. (2012). Exploring Factors Affecting Employees Job Satisfaction at Work.
Journal of Management and Social Sciences,, 8(1), 31-39.
Randhawa, K. L. (2018). Credit card fraud detection using AdaBoost and majority voting. IEEE access, 6, 14277-14284.
Sahin, Y. a. (2011). Detecting Credit Card Fraud by Decision Trees and Support Vector Machines. Proceedings of
International Multi-Conference of Engineers and Computer Scientists (IMECS 2011). Hong Kong: Vol. 1, pp.
1- 6, ISBN: 978-988-18210-3-4, ISSN: 2078-0966 (Online).
Shen, A. T. (2077). Application of classification models on credit card fraud detection. International conference on
service systems and service management (pp. 1-4). IEEE.
Terera, S.R., & Ngirande, H. (2014). The Impact of Rewards on Job Satisfaction and Employee Retention.
Mediterranean Journal of Social Sciences, MCSER Publishing,Rome-Italy., 5(1), 2039-2117.
Terera, S.R., & Ngirande, H. (2014). The Impact of Rewards on Job Satisfaction and Employee Retentions.
Mediterranean Journal of Social Science, 5(1), 2039-2117.
Tessema, M. and Soeters, J. (2006). Challenges and prospects of HRM in developing countries: testing the HRM-
performance link in Eritrean civil service. International Journal of Human Resource Management, 17(1), 86-
105.
Tiwari, P. M. (2021). Credit Card Fraud Detection using Machine Learning: A Study. arXiv preprint arXiv, 2108.10005.
Uchhana, N. R. (2021). Literature Review of Different Machine Learning Algorithms for Credit Card Fraud Detection.
International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075,
Volume-10 Issue-6.
Vikram Jeet, Dr Sayeeduzzafar. (2014). A study of HRM practices and its impact on employees job satisfaction in
private sector banks: A case study of HDFC Bank. International Journal of Advance Research in Computer
Science and Management Studies, 2(1), 62-68.
Zareapoor, M. &. (2015). Application of credit card fraud detection: Based on bagging ensemble classifier. Procedia
computer science, 48, 679-685.

59