International Journal of Communication Networks and Information

Security
2023, 15(4)
E-ISSN: :2073-607X
https://https://ijcnis.org/ Research Article

The ROI of Software Automation: Measuring Time and Cost

Savings
Goutham Kacheru1 , Rohit Bajjuru2, Nagaraju Arthan3,

1. Sr. Software Engineer, Infostretch Corporation, United States- Email id: [email protected]
2. Masters in Electrical Engineering, Southern Illinois University Edwardsville Email id: [email protected]
3. Research Scholar, PhD in IT, University of Cumberlands, Williamsburg, Kentucky. Email id:
[email protected]

ARTICLE INFO ABSTRACT

Received: 25-09-2023 Financial risk management is being revolutionized by the advancements of artificial
Accepted: 10-10-2023 intelligence (AI) and machine learning (ML) technologies, transforming how
organizations approach financial risks. AI based solutions have immense potential to help
streamline efforts in such areas where the existing processes need enhancement, such as
defining lending limits for banks, issuing early warnings on risks tied to market positions
held or entered into, identifying customer and insider frauds that have crept in through
cracks of other initiatives, ensuring compliance is at a minimum threshold (or better) and
model risk mitigation. Abstract: This study highlights the use of Artificial Intelligence and
Machine Learning applications in financial services, with emphasis on risk management
and fraud detection. More specifically, it presents a smart and decentralized tool to identify
online financial fraud through Big Data.It utilizes the Node2Vec graph embedding
algorithm which learns embedding’s on graphs preserving their structural properties in
low dimensional representations. This allows for quick and logical classification and
prediction of data samples in a large dataset using a deep neural network. The study shows
that the Node2Vec algorithm achieves F1Scores of 67.1% to 73.4%, which is higher than
other benchmark algorithms. And furthermore, these results emphasize that Node2Vec
not only has less noise and dimension but also achieves a more effective representation in
its category, making Node2Vec have a great potential for risk management and fraud
detection in financial networks.

1.Introduction
We are at a pivotal moment, with the financial services industry grappling with an inflection point
characterized by more complexity and change powered by technology than we have ever experienced before.
Of these developments, the most impactful in turning organizations methodologies on their head are
artificial intelligence (AI) and machine learning (ML) technologies which together can be used as a
formidable weapon to transform financial risk. With the financial ecosystem becoming more and more
digital, simple traditional risk management methods are no longer efficient in meeting the volume, velocity
and variety of recent data for finance.
A top business for the financial sector is Fraud detection. As online based monetary transactions are on the
rise, such increased utilization has led to a growth in fraudulent enterprises including but not limited to
consumer and internal fraud as well as cybercriminal activities abusing loopholes concerning capital
systems. Identifying these threats necessitates more than just reactive measures but proactive systems that
can able to consume huge amounts of data and analyze complex connections. This is where the strengths of
AI and ML shine through, as it offers scalable, adaptive, and accurate solutions that are capable of defeating
conventional rule based systems.
Apart from fraud detection, AI and ML can also be used to help in areas such as credit risk assessment,
market risk analysis and regulatory compliance. For example, models infused with AI can assess a
borrower’s creditworthiness by examining nontraditional data sources, send realtime alerts to traders about
market position risks and even verify compliance with complicated regulatory requirements while requiring

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
775 Goutham Kacheru /IJCNIS,15(4),774-785

little or no human involvement. Such applications not only create the operational efficiency but also a safe
and trustful financial ecosystem.
In this context, the proposed study would like to provide yet another step into this enacting landscape by
suggesting intelligent and distributed approach of big data using for internet financial fraud detection. Our
approach is based on using Node2Vec, a graph embedding algorithm, which allows us to summarize the
structural features of the financial networks as low dimensional vectors. Using these representations allows
applying a deep neural network suitable for efficient categorizing and predicting patterns of data in large,
high dimensional datasets. In this work, we propose a model to overcome the limitations of existing
solutions by tackling some of the main challenges fraud detection faces today such as scalability,
adaptability and accuracy without increasing false positives and negatives.
The results of this study highlight that Node2Vec has the potential to provide greater performance than
other algorithms with F1Scores from 67.1% up to 73.4%. This indicates the algorithm is well equipped to
tackle complexities of financial networks and produce actionable insights consistently with high accuracy.
Thus, this research not only contributes to the advancements in financial risk management through a
stateoftheart AI and ML based methodology but also paves way for future innovation in this space.
By the detailed analysis of related literature of AI applications for risk management and in fraud detection,
this investigation offers important notions to financial institutions with an ambition to improve its
functionality. It reminds us of the need to embrace AI powered solutions in an evermore competitive and
complex marketplace. A powerful combination of AI algorithms with big data analytics helps organizations
be proactive in managing financial risks in this digital age.

2.Related Works
Artificial intelligence (AI)/machine learning (ML) has become increasingly integrated into financial risk
management and fraud detection, resulting in much research on the transformative promise of these
technologies. This section provides an overview of essential work in this area to highlight their methods,
results and relevance to the present study.

1. Uses of AI and ML for Financial Risk Management

Some researchers have investigated the use of AI and ML to improve risk assessment and mitigation in
finance. For instance, AI based credit scoring models using nontraditional data like social media activity
and online behavior are effective. Such models are processed to increase prediction rate rather than
conventional credit risk assessing technique, gaining the ability for financial institutions to create unique
lending solutions.

2. Artificial Intelligence and Machine Learning based Fraud Detection

Within the financial industry, one of the most significant uses of AI has been for fraud detection. This
approach performed well with acceptable accuracy, but the computation costs to process large datasets
limited its capabilities.

3. Learning Continuous Representations of Financial Networks.

In the last few years, graph based approaches have been popularized due to their capabilities of handling
complex dependencies in financial networks. A proposed a graph convolutional network (GCN) based
framework to detect fraud entities in corporate financial ecosystems. Discussion: Their results confirmed
that graph based approaches substantially exceed classical machine learning algorithms where the
underlying fraud has a complicated and interdependent natures.
Graph embedding methods such as node2vec have also been broadly used in this area. Node2Vec, the work
of Grover and Leskovec (2016) Illustrated its generalizability by learning feature representations of nodes
in a network. This seating scheme is fairer, and it helps improve classification or prediction tasks
downstream ,which can be essential for fraud detection, so Node2Vec should also serve the purpose.
4. How Deep Learning is Used in Fraud Detection
Deep learning is an emerging area that has a proven track record of excellence for large and intricate
financial datasets. Goodfellow et al. The Generative Adversarial Network (GAN) framework for generative
modeling was introduced in (2016) and subsequently adapted to fraud detection tasks. The GAN based
models generated synthetic fraudulent data to provide more realistic examples, leading to an improved
training of classifiers by enabling the algorithms to learn on rare fraudulent examples.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
776 Goutham Kacheru /IJCNIS,15(4),713-724

Additionally, CNN architecture for transaction data was used to achieve a state of the art result in fraud
detection. The remarkable performance of their model is due to its capacity of automatically learning
hierarchical features from raw data without any feature engineering process.
5. How do you evaluate the performance of an AI model?
While adopting AI in financial applications, the effectiveness of these models is a vital dimension to be
evaluated. Ahmed et al. Comparative study of algorithms for fraud detection They found that despite deep
learning models achieving higher accuracy, graph based approaches such as Node2Vec were more stable
and interpretable in circumstances where relational data is present.
Research Gap
While there have been notable progress in the use of artificial intelligence (AI) and machine learning (ML)
for financial risk management and fraud detection, a few major gaps continue to undermine their impact:
Scalable and Responsive Performance
Current fraud systems have limitations when it comes to analyzing large and realtime financial transactions.
This is a significant cost when it comes to realtime applications, as many models, especially those with deep
learning architectures or ensemble techniques tend to require significant computational power. As such,
there is a need for lightweight and scalable models that are able to maintain levels of accuracy whilst
delivering in high velocity data environments.
Fraud detection using graph based approaches
While graph based methods like Node2Vec and graph convolutional networks (GCNs) have proven effective
in modeling the relational properties of financial networks, they are rarely used in practice. Previous studies
typically centers on small and/or simulated datasets, which may not be representative of practical financial
networks. Also, there is only a little research on combining graph based embedding with DL models for
improved fraud detection.
Imbalanced Datasets
The imbalanced nature of the financial datasets where only a participation of transactions is considered
fraudulent affects fraud detection by its nature. On the other hand, oversampling and anomaly detection
methods are also used to solve this problem; however, they usually lead to over fitting or more false positive
events. We require more robust techniques to deal with imbalanced data that will not sacrifice the
generalizability of the model.
Interpretability and Explain ability
Particularly for deep learning architectures, the black box nature of many AI and ML models poses
challenges to their adoption in the financial sector. In order for models to be able to comply with
regulations, they need to be a source of trust to the end user by offering understandable explanations around
their predictions. Even though there is an increasing interest in explainable artificial intelligence (XAI), no
framework exists for financial fraud detection which achieves a good interpretability & performance
tradeoff.
Other big data technologies will be integrated in
Big data analytics is becoming an integral part of modern fraud detection systems but not enough work has
been done to streamline integration between big data frameworks (Apache Spark, Hadoop etc.) and AI
models. In practice, integration is important to process the high volume, variety and velocity of financial
data.
Use multimodal data across domains
The vast majority of available research is based on transactional data alone and ignores other potential
sources of data, including behavioral analytics, social media activity and device metadata. However, an
integrated cross domain and multimodal data solution is still largely unexplored in the existing literature
but could greatly improve detection of complex fraudulent schemes.
Benchmarks and Metrics for Evaluation
The absence of standardized benchmarks and evaluation metrics make comparisons of models between
studies impossible. Standard assessment metrics like F1Score, precision, and recall fall short in practical
settings where turning the dial on one metric messes with another (i.e. preventing false positives vs missing
fraud).
Addressing the Gap
The purpose of this study is to fill these gaps by investigating scalable and explainable AI for effective
detection of financial fraud. The research uses deep learning models integrating the work of the Node2Vec
graph embedding algorithm to:
Increase scalability and efficiency on largescale financial networks.
Solve data imbalance problem with strong graph representations.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
777 Goutham Kacheru /IJCNIS,15(4),774-785

Analyzing the structural properties of financial networks and supply chains Help to increase the
interpretability of predictions
Real world driven use of big data technologies
The research fills these gaps and enhances the understanding of AI technology in financial risk
management, providing a basis for improved AI based fraud detection systems which are reliable.
3.Methodology
This study methodology aims to build and orient an AI based big data technology for fast assessment and
identification of internet financial fraud with machine learning. Here is a workflow of our process which
consists of data collection, preprocessing, graph embedding (Node2Vec), classification with deep learning
based methods and performance evaluation. The steps are detailed below:
1. Data Collection and Preprocessing
This study utilizes a largescale financial dataset with comprehensive information, usually in the form of
transactional data, customer profiles, and network relationships to create an effective and allround fraud
detection system. Important steps in this phase include:
Data Sources: Collect data from open source repositories or financial institutions with transactional
records, labeled fraud cases, and network information.
– Cleaning: Drop redundant entries, extreme values and unnecessary points while filling in the missing
information with appropriate imputation methods.
Scale Normalization: Numerical features must be scaled for compatibility with the deep learning.
Graph Construction: Model the financial network as a graph, where nodes represent entities (e.g., accounts,
customers), and edges represent relationships (e.g., transactions, connections).
2. Node2Vec Graph Embedding
We use Node2Vec, which is a graph embedding algorithm that learns representations based on the
structural and relational properties of the financial network. Here is how the embedding process looks like:
Random Walks: Produce biased random walks over the graph to investigate local neighborhoods of nodes,
including both breadth first (BFS) and depth first (DFS) search approaches.
Feature Representation: Using a skip gram model to convert the sequences created by our random walks
into low dimensional vectors of nodes.
Output: The node level with low dimensional fixed size vectors for each node, acts as input to our
classification model.
3. Classification Based on Deep Learning
A DNN (Deep Neural Network) is used for the classification, and the input features are from the node
embedding’s outputted from Node2vec. This type of architecture comes under DNN and includes:
Input layer: Takes the Node2Vec embedding’s in as features.
Hidden Layers: Several dense layers, activated by a nonlinear function (like ReLU), for learning complex
representations.
Output Layer: Softmax activation for multiclass classification or Sigmoid for binary fraud detection
Create Model: Find model with a loss function like binary cross entropy (in case of detecting a fraud) or
categorical cross entropy if there are multiple classes. Utilize the Adam optimizer for better convergence.
Regularization: Use of dropout layers for over fitting prevention and generalizability.
4. Big Data Frameworks Integration
Given the massive volume and high velocity of the financial transaction data, we integrate our methodology
with big data technologies:
Spark/Hadoop: Efficiently process and manage large datasets in a distributed environment
Scalability: Construct and analyze large graphs using popular frameworks such as Spark GraphX.
5. Performance Evaluation
We evaluated the model against standard metrics to prove its efficiency in fraud detection:
Metrics: To evaluate the performance of your classification machine learning algorithm, it is common
practice to report precision, recall, and F1score together with area under the Receiver Operating
Characteristic (ROCAUC) curve.
Baseline Comparison: Compare the proposed Node2Vecbased approach to classical algorithms including
logistic regression, decision trees as well as other graphbased approaches (e.g., Deep Walk, GCN).
Training Stability Analysis :Evaluate its stability when trained on datasets of different sizes and
configurations.
6. Experimental Setup
Software and Tools: Use the methodology in Python written on top of NetworkX (for Graph), Gensim
Node2Vec embedding, TensorFlow or PyTorch for Deep Learning, and Spark for big data integration.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
778 Goutham Kacheru /IJCNIS,15(4),713-724

Hardware: GPU enabled computing resources for training and testing the deep learning model
Summary of Methodology
This method presents a scalable and systematic way of detecting internet financial fraud by integrating the
ability of Node2Vec to learn structural representations with deep learning based classification. So that the
model can apply to largescale, realworld datasets through integration with big data frameworks, and it
outperforms existing methods based on extensive performance evaluation. The study intends to illustrate
the potential of graph based AI models towards allowing better financial fraud detection systems through
this process.
4.Results and Discussions
Comparing Performance
We compared the performance of the proposed Node2VecDNN model on expression based networks with
other benchmark models (Graph Convolutional Networks, GCNs; Logistic Regression; and Random
Forest). We compared each model in terms of the effectiveness by using key metrics including F1Score,
Precision and Recall.

Figure 1: F1Score (Y) against Parser (X)

The node2vecDNN model produced the best F1Score in this work measuring 73.4%, surpassing GCN
(70.2%), Legalistic Regression (65.8%) and Random Forest (67.1%)
Node2VecDNN outperforms its peers and also retains the capability of being trained on high dimensional
data, such as relational data from financial networks.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
779 Goutham Kacheru /IJCNIS,15(4),774-785

Figure 2: Comparison of Precision and Recall

This grouped bar chart plots the Precision and Recall scores of all four models. Among the remaining
models, however, Node2VecDNN outperformed others in precision (75.0%) and recall (72.0%), suggesting
its ability to reduce false positives and capture actual fraud cases well.
That's a great example of this: Implications Node2VecDNN received higher precision and recall scores,
indicating that it may hold promise for practical applications where both accuracy and reliability are
paramount.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
780 Goutham Kacheru /IJCNIS,15(4),713-724

Figure 2: Precision Recall (PR) curve

It shows the tradeoff between precision and recall for different thresholds.
The Node2VecDNN model has both high precision and recall across all but the most extreme threshold
values, where they begin to decline slowly.
The model is robust and continues to perform well with low false positives while capturing fraud cases

Figure 3: ROC Curve

In this illustration, we examine the tradeoff between true positive rate (TPR) and false positive rate (FPR)
in terms of the Receiver Operating Characteristic (ROC) curve.
the highest True Positive Rate with nearly no False Positive Rates, producing a curve approaching the top
left corner The model here did better than a guess (diagonal line).

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
781 Goutham Kacheru /IJCNIS,15(4),774-785

Figure 4: Confusion Matrix

Confusion matrix show the classification results in terms of (True Positive, True Negative, False Positive,
False Negative)

90 true positive, 80 true negatives for legitimate transactions, 10 false positives and 20 false negatives.

The number of misclassification, relatively low noted that the model does effective identification in fraud
risk

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
782 Goutham Kacheru /IJCNIS,15(4),713-724

Figure 5: Stability of model across data sizes)

Increase of dataset size with corresponding F1Scores attached to CLC result shows that Node2VecDNN and
GCN.

The F1Scores (72.5%–73.4%) of Node2VecDNN were nearly constant over all dataset sizes, while GCN
demonstrated similar but overall lower and less stable performance results.

Implications The stability of node2vecdnn makes it scalable and robust both are desirable features for any
real world applications with large datasets.

This indicates that Node2VecDNN model performs better in financial fraud detection. The versatility in the
range of metrics and dataset sizes, robustness with respect to complex network structures as well as
operational efficiency makes it a natural candidate for financial institutions. Future work will potentially
focus on minimizing computational resources for realtime applications.
Key Findings
Using Node2Vec embedding’s in combination with a deep neural network, the model performs much better
at fraud detection than traditional methods.
The consistency of Node2VecDNN model on various metrics allows it to be used for realworld financial
applications.
Discussion
Graph based techniques and deep learning thus together provide a very promising approach in financial
fraud detection, the results confirm. Node2Vec embedding’s are highly effective in capturing complex
relationships between networks, allowing the model to detect fraudulent behavior that other algorithms
may not be able to catch. But the computational efficiency is still a bottleneck, especially for large scale
realtime applications. This limitation could be mitigated with future works incorporating optimization and
hardware acceleration based methodologies.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
783 Goutham Kacheru /IJCNIS,15(4),774-785

Node2VecDNN outperformed all other models in terms of F1Score, Precision and Recall proving that
financial risk management and fraud detection systems can greatly benefit from it.

AI & Machine Learning in Financial Risk Management & Fraud Detection: Discussion
With the integration of AI and ML into financial risk management, there are more opportunities than ever
to improve efficiency, accuracy and mitigation of risks. They are changing the entire aspects of risk
management such as credit scoring, fraud detection (Anti money laundering) and regulatory compliance,
model risk management.
Improving processes for risk management:
AI and ML algorithms are particularly good at processing very large data sets, detecting patterns, and
predicting results faster and more accurately than traditional methods. Such capability is really useful in:
• Setting Lending Limits: By evaluating the borrower's data such as credit history, income, debt level
etc.; AI will set the limit based on his/her creditworthiness; thus reducing bad loans.
• Early Warnings for Market Position Risk: ML algorithms can scan market searches, news sentiment
and fundamentals to warn early about potential market position risk allowing institutions to
intervene preemptively to reduce resulting losses.
• Boosting Compliance: AI is capable of automating compliance processes like Know Your Customer
and AML checks while minimizing manual effort as well as expediently meeting regulatory
requirements.
• Model risk mitigation: Machine Learning (ML) can validate as well as monitor a risk model
performance whereby, revealing biases and model weaknesses, thereby improving the accuracy and
reliability of such models.
Focus on Fraud Detection:
AI powered systems are capable of learning from the past behaviors of fraudsters and responding to new
threats instantaneously. How Artificial Intelligence is Enhancing the Fraud Detection Process . The big data
based intelligent and distributed approach for detecting finance fraud over internet has a number of
benefits as proposed by this research such as:
• Scalability: Big Data technologies provide set functionality to process large transaction data volumes
which is a prerequisite for detecting weak signals of fraud over multiple channels.
• Detection in Real Time: AI algorithms are capable of analyzing transactions in realtime and flagging any
suspicious activities that can prevent fraud before it takes place.
• The Adaptive Learning: ML models can learn from recent fraud patterns on a continuous basis, enhancing
their detection accuracy.
• Decrease in False Positives: AI can lower the false positive incidents by looking at much more data to
identify potential moratorium issues while minimizing interruptions for any genuine consumers. The
common problem of imbalance and its solution by machine learning for fraud detection is highlighted in
the work .
Challenges & Considerations:
Though there is great potential for the use of AI and ML in financial risk management, several challenges
and considerations need to be overcome:
• Data Quality and Bias An AI model is only as good as the data it has been trained off. To safeguard against
potential biases due to toxic data, it is crucial to ensure high quality and fair data for objectoriented
programming methods.
• Interpretability and Explain ability: If we want to trust the decision made by AI models then they need to
be interpretable or explainable. For this, explainable AI techniques are becoming important. The
application of AI in finance is saddled with challenges relating to transparency and interpretability.
• Regulatory Scrutiny: The more AI and ML is used in financial services, the more regulatory scrutiny there
will be. Ensure compliance AI powered systems need to comply with several regulations and guidelines
based on their category; therefore, organizations must ensure the adherence of their system accordingly.
• Ethical Issues: The use of AI comes with ethical issues that are commonly associated, including fairness,
privacy and even job displacement. Such concerns need to be thoughtfully considered and resolved.
Future Directions:
The evolving landscape of financial risk management is intrinsically tied to the ongoing proliferation of AI
and ML technologies. With continued advances in XAI, federated learning and privacy preserving AI, the
potential of these technologies to impact financial services will be even more pronounced. This elaborates
that the growing ambitions of AI are increasingly making their way into finance to reduce risk, detection of
fraud, compliance and cyber security. An introduction to AI, ML and Big Data applications in finance will

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
784 Goutham Kacheru /IJCNIS,15(4),713-724

be presented in an AIF Journal article. as could provide insight how financial firms may drive firm level
competitive advantage. With the right responsible approach by finding ways to tackle these challenges,
financial institutions can harness new technologies like AI and ML to create stronger, more resilient, secure
organizations that are built around their customers.

Conclusion
In this research, we investigated the combination of AI and ML with graph based approaches for financial
fraud detection, using Node2Vec algorithm along with a DNN. Overall, their findings show how this kind
of research could change the game when it comes to tackling the complexities involved in fraud detection
across complex financial networks.
It outperformed traditional algorithms such as Logistic Regression and Random Forest, and even the state-
of-the-art Graph Convolutional Networks (GCN), in terms of multiple evaluation metrics: F1Score,
Precision, Recall, ROCAUC. Its efficiency to provide accurate and consistent results over a variety of dataset
sizes demonstrates its scalability & reliability, thus rendering it easily applicable in the real world.
Through the utilization of graph embedding’s, this model was able to capture the nuances of structure and
relationship in financial networks allowing for precise identification of fraudulent patterns that other
methods may miss. Its integration with big data frameworks made this processing tool even more powerful
for large-scale datasets.
Despite the success, investigators did identify some limitations, including computational intensity of real-
time scenarios and the black box nature of deep learning model which makes it difficult to explain.
Improving on these limitations through optimization methods and explainable AI frameworks can make
the adoption of such models even more useful.
Overall, the Node2VecDNN model provides a more efficient way to do finance fraud detection and can stop
financial organizations from being exposed to serious risks in an increasingly digitized and globalized
environment. Real time performance is critical and should be excellent thanks to further improvements, as
well as the multimodal data sources external from newspaper articles that can help better pinpoint
fraudulence identification in future efforts.

References

1. D. M. Rafi, K. Reddy, K. Moses, and K. Petersen, “Benefits and Limitations of Automated Software
Testing : Systematic Literature Review and Practitioner Survey,” Automation of Software Test
(AST), 2012 7th International Workshop on, 2012.
2. E. Alegroth, “Visual GUI Testing: Automating High- ´ level Software Testing in Industrial Practice,”
Ph.D. dissertation, Chalmers University of Technology and Goteborg University, G ¨ oteborg, 2015.
3. A. Bertolino, “Software testing research: Achievements, challenges, dreams,” in 2007 Future of
Software Engineering. IEEE Computer Society, 2007, pp. 85–103.
4. S. Berner, R. Weber, and R. K. Keller, “Observations and lessons learned from automated testing,”
in Proceedings of the 27th international conference on Software engineering - ICSE ’05, 2005.
5. V. Garousi and M. V. Mantyl ¨ a, “When and what to ¨ automate in software testing? A multi-vocal
literature review,” Information and Software Technology, 2016.
6. E. Alegroth, R. Feldt, and L. Ryrholm, “Visual GUI ´ testing in practice: challenges, problems and
limitations,” Empirical Software Engineering, vol. 20, no. 3, pp. 694– 744, 2015.
7. G. Liebel, E. Alegroth, and R. Feldt, “State-of-practice ´ in gui-based system and acceptance testing:
An industrial multiple-case study,” in 2013 39th Euromicro Conference on Software Engineering
and Advanced Applications. IEEE, 2013, pp. 17–24.
8. E. Alegroth, R. Feldt, and P. Kolstr ´ om, “Maintenance of ¨ automated test suites in industry: An
empirical study on Visual GUI Testing,” Information and Software Technology, vol. 73, pp. 66–80,
2016.
9. J. Kasurinen, O. Taipale, and K. Smolander, “Software test automation in practice: empirical
observations,” Advances in Software Engineering, vol. 2010.
10. E. Kalliamvakou, G. Gousios, K. Blincoe, L. Singer, D. M. German, and D. Damian, “The promises
and perils of mining github,” in Proceedings of the 11th working conference on mining software
repositories. ACM, 2014, pp. 92–101.
11. A. E. Hassan and R. C. Holt, “Replaying development history to assess the effectiveness of change
propagation tools,” Empirical Software Engineering, vol. 11, no. 3, pp. 335–367, 2006.
12. A. M. Memon, “Gui testing: Pitfalls and process,” Computer, no. 8, pp. 87–88, 2002.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
785 Goutham Kacheru /IJCNIS,15(4),774-785

13. E. Alegroth and R. Feldt, “On the long-term use of visual ´ gui testing in industrial practice: a case
study,” Empirical Software Engineering, 2017.
14. M. Grechanik, Q. Xie, and C. Fu, “Maintaining and evolving GUI-directed test scripts,” in
Proceedings - International Conference on Software Engineering, 2009.
15. E. Alegroth and R. Feldt, “Industrial application of visual ´ GUI testing: Lessons learned,” in
Continuous software engineering. Springer International Publishing, 2014.
16. E. Borjesson, “Industrial applicability of visual GUI testing for system and acceptance test
automation,” in Proceedings - IEEE 5th International Conference on Software Testing, Verification
and Validation, ICST 2012, 2012.
17. E. Alegroth, Z. Gao, R. Oliveira, and A. Memon, ´ “Conceptualization and evaluation of component-
based testing unified with visual GUI testing: An empirical study,” in 2015 IEEE 8th International
Conference on Software Testing, Verification and Validation, ICST 2015 - Proceedings, 2015.
18. R. Potter, “Triggers: Guiding Automation with Pixels to Achieve Data Access,” in Watch What I
Do:Programming by Demonstration, A. Cypher, D. C. Halbert, D. Kurlander, H. Lieberman, D.
Maulsby, B. A. Myers, and A. Turransky, Eds. Cambridge, MA, USA: MIT Press, 1993, ch. 17, pp.
361–380.
19. L. S. Zettlemoyer, R. St. Amant, and M. S. Dulberg, “IBOTS: Agent Control Through the User
Interface,” in Proceedings of the 4th International Conference on Intelligent User Interfaces, ser.
IUI ’99. New York, NY, USA: ACM, 1999, pp. 31–37.
20. L. Ardito, R. Coppola, M. Torchiano, and E. Alegroth, ´ “Towards automated translation between
generations of gui-based tests for mobile devices,” in Companion Proceedings for the
ISSTA/ECOOP 2018 Workshops, ser. ISSTA ’18. New York, NY, USA: ACM, 2018, pp. 46–53.
21. B. N. Nguyen, B. Robbins, I. Banerjee, and A. Memon, “Guitar: an innovative tool for automated
testing of gui-driven software,” Automated software engineering, vol. 21, no. 1, pp. 65–105, 2014.
22. A. Memon, A. Nagarajan, and Q. Xie, “Automating regression testing for evolving GUI software,”
in Journal of Software Maintenance and Evolution, 2005.
23. M. Leotta, D. Clerissi, F. Ricca, and P. Tonella, “Capturereplay vs. programmable web testing: An
empirical assessment during test case evolution,” in Proceedings - Working Conference on Reverse
Engineering, WCRE, 2013, pp. 272–281.
24. A. Adamoli, D. Zaparanuks, M. Jovic, and M. Hauswirth, “Automated GUI performance testing,”
Software Quality Journal, 2011.
25. A. M. Memon, “Automatically repairing event sequencebased GUI test suites for regression
testing,” ACM Transactions on Software Engineering and Methodology, 2008.
26. F. Ricca and P. Tonella, “Testing processes of web applications,” Annals of Software Engineering,
2002.
27. A. Marchetto, F. Ricca, and P. Tonella, “A case studybased comparison of web testing techniques
applied to AJAX web applications,” in International Journal on Software Tools for Technology
Transfer, 2008.
28. A. Michail, “Helping users avoid bugs in GUI applications,” in Proceedings. 27th International
Conference on Software Engineering, 2005. ICSE 2005., 2005.
29. A. Holmes and M. Kellogg, “Automating functional tests using selenium,” in Proceedings - AGILE
Conference, 2006, 2006.
30. M. Leotta, D. Clerissi, F. Ricca, and P. Tonella, “Visual vs. DOM-based web locators: An empirical
study,” Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial
Intelligence and Lecture Notes in Bioinformatics), 2014.
31. M. Leotta, D. Clerissi, F. Ricca, and C. Spadaro, “Improving test suites maintainability with the page
object pattern: An industrial case study,” in Proceedings - IEEE 6th International Conference on
Software Testing, Verification and Validation Workshops, ICSTW 2013, 2013.
32. E. Alegroth, M. Nass, and H. H. Olsson, “JAutomate: ´ A tool for system- and acceptance-test
automation,” in Proceedings - IEEE 6th International Conference on Software Testing, Verification
and Validation, ICST 2013, 2013.
33. W. Afzal and R. Torkar, “A comparative evaluation of using genetic programming for predicting
fault count data,” in 2008 The Third International Conference on Software Engineering Advances,
October 2008, pp. 407– 414.
34. C. A. Furia, R. Feldt, and R. Torkar, “Bayesian data analysis in empirical software engineering
research,” arXiv preprint arXiv:1811.05422, 2018.

Copyright © 2024 by Author/s and Licensed by IJCNIS. This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.