MACHINE LEARNING CLASSIFIERS TO ANALYZE CREDIT RISK

© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1412
MACHINE LEARNING CLASSIFIERS TO ANALYZE CREDIT RISK
Jyoti Tiwari1, Pratiti2, Pragalbh Mishra3, Sarthak Tiwari4, Shyam Dwivedi5
1,2,3,4Student of B. Tech Final Year, Dept. of Computer Science engineering, Rameshwaram Institute of Technology and
Management, Lucknow
5Assistant Professor & Head of Department, Dept of Computer Science engineering, Rameshwaram Institute of
Technology and Management, Lucknow
--------------------------------------------------------------------------***----------------------------------------------------------------------------
Abstract: Corporate bankruptcy has the potential to
devastate the economy. A multinational business bankruptcy
can upset the global financial ecosystem, as an increasing
number of corporations expand internationally to benefit on
foreign resources.
Corporations do not fail overnight; objective metrics and a
thorough examination of qualitative (e.g. brand) and
statistical (e.g. econometric components) data may assist in
determining a company's financial risk. With recent
improvements in communication and information
technology, gathering and storing data about a company
has grown easier. The primary operation of the banking
business is lending money to individuals in need. The
depositor bank collects the interest paid by the principle
borrowers in order to repay the principle borrowed from the
depositor bank. In the subject of financial risk management,
credit risk analysis is becoming increasingly significant. The
credit risk of the customer dataset is assessed using a variety
of credit risk analysis approaches. The challenging work of
evaluating credit risk datasets to determine whether to
grant the client a loan or reject his or her application is a
demanding undertaking that requires a thorough
examination of the customer's credit dataset or data.
Because of its effectiveness in learning complex models,
machine learning has become a prominent subject in big
data analytics in recent years. Support vector machines,
adaptive boosting, artificial neural networks, and Gaussian
processes may all be used to recognise patterns in data that
humans would miss. This study examines several credit risk
analysis strategies that are used to assess credit risk
datasets.
Key Words: Machine learning, Credit risk, Financial
Risk, Effectiveness, Gaussian processes, Support vector
machines, artificial neural networks,
1. INTRODUCTION
Credit scoring systems are an important aspect of a
company's managing risk since they detect, analyse, and
track consumer credit risk (Brigham, 1992; Johnson &
Kallberg, 1986). Customers are allocated to risk classes
based on their particular propensities to default on
payments to assess the default risk associated with loan
sales. The default probability can be derived either
externally or internally using a scoring model. The primary
internal source of creditworthiness data is a company's
accounting department, which may give information on a
customer's prior payment history as well as personal
attributes like age, education, career, and domicile.
Companies can also turn to commercial credit agencies,
which gather information on consumers based on criteria
including delinquent invoices, court-ordered payment
demands, enforcement proceedings, and uncovered
checks. Applicants with poor financial standing have a
limited likelihood of repaying the loan, and hence are
defaulters. Different forms of credit risk evaluation
algorithms are utilised to decrease the defaulter's rate in
the credit dataset. Even a slight improvement in credit
evaluation accuracy can sometimes result in large losses
being avoided. The methods that are utilised to make these
judgments will be the topic of this study. Algorithms are
employed in a variety of disciplines to achieve various
goals. For example, they are employed in businesses to
Recruit people who fit the proposed profile. Algorithms
can make the procedure easier and faster and more fluid,
for example Algorithms, on the other hand, are indeed a set
of codes having specific goals in mind objectives. It might,
for example, establish prejudice or a special preference
throughout the recruiting process profile and then
"format" the people who work for the company.
Transparency is essential in this modern digital and Big
Data era; it should be one that enables for transformation
and advancement rather than hindering it. This field's
words must be ethical, transparent, well-known, and easy
to understand. To help achieve these goals, data specialists
must be trained on how to apply machine learning
algorithms, as well as their limits. This study is unique in
that it addresses certain specific concerns that arise when
using Big Sophisticated algorithms. We primarily address
issues concerning the application of algorithms to solve or
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072

achieve a goal. We use machine learning techniques to
predict the needed default probability for a large set of
data of short-term instalment credits in this article. To get
consistent probabilities, we apply the following method.
Because of the large number of decisions that must be
made in the consumer lending industry, it is necessary to
rely on models and algorithms rather than human
discretion, and to base such algorithmic decisions on
"hard" data, such as characteristics found in consumer
credit files collected by credit bureau agencies.
2. METHODLOGY
For a more accurate and trustworthy credit risk analysis,
many methodologies are employed in the evaluation of
credit datasets. In We've examined many ways in this
literature review. to the evaluation of credit risk
A. BAYSIAN CLASSIFIER
The notion behind a Bayesian classifier is that the purpose
of a (natural) category is to anticipate the values of
characteristics for its individuals. Because the
characteristics have similar values, the examples are
divided into classes. Natural types are a term for such
classes. The goal feature in this section is a discrete class
that is not strictly binary. A Bayesian classifier works on
the principle that if an agent understands the category, it
can anticipate the results of the other attributes. Bayes'
rule may be used to forecast the class given (part of) the
feature values if the class is unknown. The learning agent
in a Bayesian classifier creates a probabilistic model of the
characteristics and uses it to anticipate the categorization
of a new instance.
BAYES THEOREM FORMULA
P(A/B)=P(A⋂B)/P(B) = P(A).P(B/A)/P(B)
Where:
P(A) = The probability of A occurring
P(B) = The probability of B occurring
P(A∣B) =The probability of A given B
P(B∣A) = The probability of B given A
P(A⋂B)= The probability of both A and B occurring
B. NAÏVE-BAYE CLASSIFIER
The Bayes' Theorem-based Naive Bayes classifiers are a
group of classification algorithms. It is a group of
algorithms that share a similar premise, namely that each
pair of characteristics being categorized is independent of
the others. Consider the following dataset as an example.
Consider a hypothetical dataset that represents golfing
weather. Each tuple assigns a fit("Yes") or unfit("No")
rating to the meteorological circumstances.
C. DECISION TREE
A decision tree is a decision-making aid that employs a
tree-like representation of decisions and their probable
outcomes, such as chances event results, capital
resources, and utility. It's one way of displaying a
conditionally control algorithm.
Decision tree algorithm are used in operations research,
particularly in decision analysis, to assist determine the
best method for achieving a goal, but they are also a
popular machine learning technique.
D. K-NEAREST NEIGHBOR
The KNN method is a basic supervised machine learning
technique that can address both classiﬁcation issues. It's
simple to set up and operate, but it has the disadvantage
of being noticeably slower as the amount of data in use
rises. K-Nearest Neighbor is a Supervised Learning-based
Machine Learning algorithm that is one of the most basic.
The K-NN method assumes that the new case/data and
existing cases are comparable and places the new case in
the category that is most similar to the existing categories.
The K-NN method maintains all of the available data and
classifies a new data point based on its resemblance to the
existing cheval. This implies that fresh data may be

quickly sorted into a suitable category using the K-NN
method. The K-NN algorithm may be used for both
regression and classification, however it is more
commonly utilized for classification tasks. The K-NN
algorithm is a non-parametric algorithm, therefore means
it makes no assumptions about the data. It's also known as
a lazy learner algorithm since it doesn't gain from the
training set right away; instead, it saves the dataset and
uses it to classify. The square root of the sum of the
squared differences between the two vectors is used to
determine Euclidean distance.
E. PERCEPTRON WITH MULTILAYERS: -
MLPs are commonly utilized in the banking sector to
manage credit risk. For supervised learning, machine
learning uses the back propagation technique. It consists of
an input layer, an output layer, and one or even more
hidden layers in between. All of the levels are totally
interconnected. The nodes, which act like neurons, are the
processing elements of each layer except the input layer.
Each node in one layer is connected to another node in the
next layer, which is coupled to a set of weights. Neurons
with nonlinear activation function exist in numerous
layers. The network can learn the connection among input
and output vector thanks to these layers.
F. VECTOR MACHINE SUPPORT: -
SVM (Support Vector Machine) is a common Artificial
Learning technique for Classification and Regression.
However, it is most commonly employed in Machine
Learning for Classification issues. The SVM algorithm's
purpose is to find the optimal line or decision boundary
that can divide n-dimensional space into classes so that
fresh data points may be readily placed in the proper
category in the future. A hyperplane is a term for the
optimal decision boundary. SVM selects the hyperplane-
helping extreme points/vectors. Support vectors are the
extreme examples, and the technique is named after them.
3. NEURAL NETWORK MAN-MADE: -
The concept of neural networks started with
'connectionism,' a model of how neurons in the brain
function that employed linked circuits to imitate intelligent
behaviour. Neurophysiologist Warren McCulloch and
mathematician Walter Pitts depicted it with a simple
electrical circuit in 1943. In his book The Organization of
Behaviour (1949), Donald Hebb developed the concept
further, stating that neural networks strengthen with each
usage, particularly connecting neurons that fire at the
same time, thereby beginning the long trek towards
measuring the brain's complicated functions. Precursors to
Neural Networks are two important concepts.
(a)- 'Threshold Logic' is a technique for transforming
continuous input into discrete output.
(b)- 'Hebbian Learning' is a brain plasticity-based learning
paradigm presented by Donald Hebb in his book "The
Organization of Behavior," which is commonly stated as
"Cells that fire along, wire around each other."
Both ideas were offered in the 1940s. The first Hebbian
network was successfully implemented at MIT in 1954, as
academics attempted to convert these networks onto
computing systems in the 1950s.
4. BUILD THE CLASSIFIER:-
The term "ensemble" comes from the Latin meaning
"union of pieces." The most often used classifiers are prone
to making mistakes. These mistakes are unavoidable, but
they may be minimised with effective learning classifier
building.
Ensemble learning is a method of constructing many base
classifiers from which a new classifier is created that
outperforms all constituent classifiers. The method, hyper

parameters, representation, and training set may differ
amongst these base classifiers.
Ensemble approaches have the primary goal of lowering
bias and variation.
A.Bagging:-
By taking samples the practise set "with substitute,"
different training sets of the same size can be created.
Breiman created the Bagging approach in 1996 as a front-
standing ensemble strategy. This is a learning algorithm
ensemble Meta method for improving the reliability and
accuracy of machine learning used in statistical
classification and regression. It aids in the reduction of
oversizing. A specific instance of the model averaging
method is bagging. Typically, similar types of classifiers are
used as foundation classifiers in bagging. By employing the
bagging method, one may generate various decision
structures.
B.Boosting:-
Many analysts misunderstand the word 'boosting' in data
science. Let me give you a fascinating explanation of this
phrase. Boosting gives machine learning models more
power to increase their prediction accuracy. In data
science contests, boosting methods are one of the most
often utilized algorithms. The winners of our previous
hackathons all agreed that they would use a boosting
technique to increase the accuracy of their models. In this
essay, I will describe the enhancing method in a
straightforward manner. Below are the Python codes. I've
avoided the Boosting's daunting mathematical derivations.
Because it would have prevented me from explaining this
subject in layman's terms.
5. USED DATABASE:-
For the execution of machine learning techniques and
ensemble learning, primarily two credit datasets, such as
the USA credit and England credit datasets, are employed.
These two datasets came from the UCI Machinery
Repository (http://archive.ics.uci.edu/ml/). The dataset
has two classes: "good" and "bad" creditors.
Table 1. summarises the credit dataset that was used
6. ANALYZE AND COMPARE: -
Table 2. shows the classifiers' accuracy.
Using the USA and England datasets, we examined the
classifier accuracies of several techniques.
7. OBJECTIVE: -
Understanding the goals of management of credit risk
might assist you as a customer even when you're not in the
banking business. Every loan that a lender considers is
subject to credit risk management. Banking institutions
must strike a difficult balance between stringent credit risk
standards and client happiness. This balance is achieved by
using conservative credit risk management practises, quick
loan approvals, and fair loan pricing to safeguard loan
portfolios while maintaining account holders pleased.
Algorithm Dataset Accuracy
Bayesian
classifier
USA England 81.21
76.16
Naive-baye
classifier
USA England 75.61
71.32
Decision tree USA England 89.90
84.41
KNN USA England 86.45
71.25
MLPs USA England 87.12
78.32
SVM USA England 83.53
77.81
Dataset Attributes Instances Classes
USA 16 760 2
England 23 1100 2

8. FUTURE SCOPE: -
We agreed that because our credit scoring model achieved
a high level of accuracy and gained the trust of
Microfinance institutions as a reliable tool, they would
tend to input even more accurate and detailed data, even
for attributes that we had to fill in with unknown values
during the pre-processing phase. This will allow us to
rebuild the model in the near future and construct a new,
improved model with the same or better accuracy. Do you
know what your credit rating is? Have you been denied
credit for no apparent reason? A credit file exists for
anybody who has ever borrowed money to apply for a
credit card, purchase a vehicle, a house, or any other
personal loan.
9. CONCLUSION:-
This study is based on a real initiative whose main goal
was to help financial companies manage credit risk. Many
issues confront financial firms. Apart from the danger of
allowing a loan to a client who may default, they have little
chance of earning by denying loans to a borrower who is
capable of meeting his commitments. As a result, several
financial institutions have adopted this strategy. Credit
scoring models, which are capable of detecting hidden
trends in very large databases and classifying consumers
as default or nondefault, have recently been introduced.
The Credit Scoring System proved to be very accurate,
identifying 100% of default consumers. It's important to
remember that the paper's goal is to survey the various
credit risk classifiers. Various types of classifiers including
ensemble classifiers are described in this study.
REFRENCES
[1] Pandey, T.N., Jagadev, A.K., Choudhury, D. and
Dehuri, S., ‘Machinelearning-based classifiers
ensemble for credit risk assessment’, Int. J.
Electronic Finance, Vol. 7(3/4), pp.227–249, 2013.
[2] Shih, K-H., Hung, H-F., Lin, B., ‘Construction of
classification models for credit policies in banks’,
Int. J. of Electronic Finance, Vol. 4(1), pp.1–18,
2010.
[3] Sokolova, M., Lapalme, G., ‘A systematic analysis of
performance measures for classification tasks’,
Journal of Information Processing and
Management, Vol. 45, pp.427–437, 2009
[4] Trilok Nath Pandey, Alok Kumar Jagadev, Suman
Kumar Mohapatra, Satchidananda Dehuri, ‘Credit
Risk Analysis using Machine Learning Classifiers’
International Conference on Energy ,
Communication , Data Analytics and Soft
Computing (ICECDS-2017)
[5] Huang, G.B., Chen, L., Siew, C.K., ‘ Universal
approximation using incremental networks with
random hidden computation nodes’, . IEEE Trans
Neural Networks, Vol. 17(4), pp. 1243-1289, 2006
[6] Jain, A., Kumar, A.M., ‘ Hybrid neural network
models for hydrologic time series Forecasting’,
Applied Soft Computing, Vol. 7 (2), pp. 585–
592, 2007.
[7] Hui, Xiang., Yang, S.G., ’Using clustering-based
bagging ensemble for credit scoring’. Business
Management and Electronic Information (BMEI),
2011 International Conference on. Vol. 3. IEEE,
2011.
[8] Dietterich, T.G., ‘Experimental comparison of three
methods for constructing ensembles of decision
trees: bagging, boosting, and randomization’,
Machine Learning, Vol. 40, pp.139–157, 2000.
[9] Huang, J., Chen, H., Hsu, C.J., ‘Credit rating analysis
with SVM and neural network : A market
comparative study’, Decision Support System, Vol.
37, pp. 543-558, 2004
[10] https://en.wikipedia.org
[11] https://www.javatpoint.com
[12] https://towardsdatascience.com
[13] https://www.google.com
[14] https://www.investopedia.com
[15] https://www.geeksforgeeks.org
[16] Bask, A., Merisalo-Ratanen, H., Tinnila, M. and
Lauraeus, T., ‘Towards e-banking: the evolution of
business models in financial services’,
International Journal of Electronic Finance, Vol.
5(4), pp. 333– 356,2011.

[17] Bekhet, H.A., Al-alak, B.A., ‘Measuring e-statement
quality impact on customer satisfaction and
loyalty’, International Journal of Electronic
Finance, Vol. 5(4), pp.299–315, 2011.
[18] Curran, K., Orr, J., ‘Integrating geolocation into
electronic finance applications for additional
security’, International Journal of Electronic
Finance, Vol. 5(3), pp.272–285,2011.
BIOGRAPHY
Jyoti Tiwari - She is currently Student of B.
Tech Final Year, Dept. of computer science
engineering, Rameshwaram Institute of
Technology and Management, Lucknow and
working on the project Machine ‘Learning
Classifiers To Analyze Credit Risk.’
Pratiti- She is currently Student of B. Tech
Final Year, Dept. of computer science
engineering, Rameshwaram Institute of
Technology and Management, Lucknow and
working on the project Machine ‘Learning
Classifiers To Analyze Credit Risk.’
Pragalbh Mishra - He is currently Student
of B. Tech Final Year, Dept. of computer
science engineering, Rameshwaram
Institute of Technology and Management,
Lucknow and working on on the project
‘Machine Learning Classifiers To Analyze
Credit Risk.’
Sarthak Tiwari - He is currently Student
of B. Tech project Machine Learning
Classifiers to Analyze Credit Risk Final
Year, Dept. of Mechanical engineering,
Rameshwaram Institute of Technology and
Management, Lucknow and working on
the project ‘Machine Learning Classifiers
To Analyze Credit Risk.’
Shyam Dwivedi - He is currently working
as an Assistant Professor and Head of
Department in Rameshwaram Institute of
Technology & Management, Lucknow,
India. He is M. Tech – 2012 BIT Mesra,
Ranchi, he has a teaching experience of 10
years and 1-year in TCS Industrial
experience

MACHINE LEARNING CLASSIFIERS TO ANALYZE CREDIT RISK

Recommended

Recommended

More Related Content

Similar to MACHINE LEARNING CLASSIFIERS TO ANALYZE CREDIT RISK

Similar to MACHINE LEARNING CLASSIFIERS TO ANALYZE CREDIT RISK (20)

More from IRJET Journal

More from IRJET Journal (20)

Recently uploaded

Recently uploaded (20)

MACHINE LEARNING CLASSIFIERS TO ANALYZE CREDIT RISK