See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/372656643

LOAN ELIGIBILITY PREDICTION USING MACHINE LEARNING: A

COMPARATIVE APPROACH

Article · January 2023

CITATIONS READS

5 6,593

4 authors, including:

Debasish Swapnesh Kumar Nayak Tripti Swarnkar

Centurion University of Technology and Management National Institute of Technology Raipur
62 PUBLICATIONS 301 CITATIONS 128 PUBLICATIONS 1,193 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Debasish Swapnesh Kumar Nayak on 27 July 2023.

The user has requested enhancement of the downloaded file.

 Global Journal of Modeling and Intelligent Computing
(GJMIC)
Vol. 3, No. 1, January, 2023 ISSN: 2767-1917

LOAN ELIGIBILITY PREDICTION USING MACHINE LEARNING: A

COMPARATIVE APPROACH

Sonali Kumari, Siksha ‘O’ Anusandhan University, India, (chaturvediswarna864@gmail.com)

Debasish Swapnesh Kumar Nayak, Siksha ‘O’ Anusandhan University, India (swapnesh.nayak@gmail.com)
Tripti Swarnkar, Siksha ‘O’ Anusandhan University, India (triptiswarnkar@soa.ac.in)

Abstract - Machine learning (ML) algorithms can bring revolution in the research field in almost all areas.
Processes in numerous industries, including finance, real estate, security, and genomics, are being transformed
by machine learning (ML) algorithms. One of the major impediments in the banking sector is the loan
approval process. Modern tools like ML models help accelerate, streamline, and increase the precision of loan
approval procedures. It will benefit both the client and the bank in terms of time and manpower required for
loan eligibility prediction. The entire work is cantered on a classification problem and is a form of supervised
learning in which it is important to determine whether the loan will be approved or not. Also, it is a predictive
modeling problem where a class label is predicted from the input data for a specific sample of input data. In
this work, we deployed various ML algorithms to identify the loan approval status and compare the
performance of implemented models. The implemented models will attempt to predict our target column on
the test dataset using information from the loan eligibility prediction dataset obtained from Kaggle, which
includes features like loan amount, number of dependents, and education. The parameters like accuracy,
confusion matrix, ROC curve, and precision are measured for specific models whose performance is
significant.

Keywords: Machine Learning (ML), Supervised Learning, Loan Eligibility Prediction (LEP), Kaggle, Real
estate.

1. INTRODUCTION
The banking sector is a crucial component of any economy. Banks serve as intermediaries between savers and
borrowers and are responsible for providing financial services to individuals, businesses, and governments.
One of the most important services provided by banks is loans. Loans are essential for the growth of the
economy (King et al., 2001). They allow individuals and businesses to invest in new ventures, buy homes and
cars, and make other purchases that they would not be able to afford without borrowing. Loans also help
businesses to expand their operations, hire more employees, and ultimately contribute to the growth of the
economy. In addition to providing access to credit, banks also play a critical role in managing risk. Banks
carefully evaluate loan applications to determine the creditworthiness of the borrower and assess the risks
associated with the loan (Samreen, 2012). By doing this, the risk of default is reduced because loans are given
to borrowers who are more likely to repay them. Banks create lending policies, and according to such
regulations, loans are approved based on the status of the applicant. Loan applications are reviewed in
accordance with the standing of the applicant under the lending policies established by banks. Loans are
frequently only approved by banks after a thorough evaluation of the applicant’s condition, either through
methodically examining submitted documents or through direct asset verification. However, there is no
guarantee that the individual who was selected from all of the candidates is the best one (Sheikh et al., 2020).
Machine learning can help automate loan eligibility prediction by analyzing vast amounts of data and
identifying patterns and trends to eliminate human error. One of the main benefits of machine learning is that
it can learn from past data and use that knowledge to predict future outcomes (Srivastava et al., 2018). By
training machine learning algorithms on historical loan data and outcomes, banks can develop models that can
accurately predict loan eligibility based on various factors such as the qualification of an applicant, gender,
employment status, and other relevant factors. The machine learning models can also help identify critical risk
factors that can be used to refine the loan application process by identifying high-risk applications and
prompting human review (Al Mamun et al., 2022). By making loan decisions more quickly and accurately,
eliminating the need for manual review, and making personalized loan recommendations, machine learning
can also enhance the overall client experience. Overall, machine learning can enhance the speed and accuracy
of loan processing while also enhancing the applicants experience (Singh et al., 2021).

48
Global Journal of Modeling and Intelligent Computing (GJMIC) ISSN: 2767-1917, Vol. 3, No.1, January 2023

In this paper, we explored the capability of various machine-learning classification algorithms. The analysis
of these algorithms leads to the suggestion of an original solution in this area. The deployment of various
models and the model architecture is shown in Figure 1.

Figure 1: Proposed pipeline of the Model

2. RELATED WORK

C. Naveen Kumar et al., (2022), discussed the traditional loan approval process and how they are unsafe for
the bank. They discussed data mining and machine learning approaches to automate the loan approval process
so that it will save time and resources. This work aims to replace the traditional loan eligibility approval
process in the banking industry.

Park et al., (2021), compared six different machine-learning algorithms based on precision, recall, and
accuracy. The objective of this work is to predict if a loan will be approved or not. Random Forest displayed
the most accuracy in their study, at 95.55%, whereas Logistic Regression displayed the lowest accuracy.

Mohankumar et al., (2016), discussed how feature selection is important in a predictive model. They used
several methods and approaches for the same. They also suggested that a linearneural network can be used
instead of a linear regression model to take advantage of the vivid powerof an artificial neural network.

Awodele et al., (2022), discussed that the cascade of Deep Learning networks and Support vector machines
improves accuracy by 9%. In this study, the authors employed DNN to convert input data from a lower
dimension to output features in a higher dimension, which wasfurther used to upskill a support vector
machine-based classification model.

Pidikiti et al. (2019), the main objective was to reduce the risk involved with choosing a reliableapplicant to
sanction the loan that will many resources including efforts and cost for the bank.In this article, they used
machine learning techniques like classification, logistic regression, decision trees, and gradient boosting to
forecast loan data. The decision tree algorithm performed best when compared to other algorithms, with an
accuracy of 82 percent. It was effective since the classification problem results were improved.

Foretelling loan non-payers is the most burdensometask for every bank, according to Pandey et al. (2010).
However, banks can drastically cut their losses by minimizing non-profit assets by foreseeing loan defaulters.
The study of loan approval prediction thus became essential. Machine learning algorithms are very significant
and helpful in the prediction of this kind of data. This study deployedfour machine learning algorithms and
later compared them. With a high accuracy of 79.67%, Support Vector Machine proved to be the most
efficient model for predicting loan acceptance in this case. From various banks that had supported several
boundary advances, they collected a list (dataset) of former clients’ information.

An assertion regarding what one anticipates will happen in the future is known as a prediction. Every day,
predictions are made. Some are serious and founded on mathematical calculations, while others are only
educated guesses. Predicting what will happen in the future, whether it be in a few months, a year, or ten
years, can help us in several ways. In their 2016 study, Kumar Arun et al. experimented the ways to predict
what are the different proceduresin which any bank can sanction loan applications. They deployed a model
that makes use of machine learning tools like neural networks and SVM.

49
Global Journal of Modeling and Intelligent Computing (GJMIC) ISSN: 2767-1917, Vol. 3, No.1, January 2023

According to the clients historical financial and credit scores, Tejaswini et al. (2020) proposed a vigorous
predictiveapproach to decide whether an applicant should be accepted or rejected. This study set out to
provide a quick, easy, and efficient method for selecting qualified applications. The dataset used was collected
from different commercial organizations. The dataset was divided into training and testing, the training
dataset was used to train the model whereas the testing dataset was used to validate the model. Three machine
learning algorithms were used in this study to forecast customer loan approval. The evaluation outcomes
suggestthat the Decision Tree algorithm has outperformed all models with the highest accuracy of 82%.

A Solid machine-learning model to forecast loan acceptance was put forth by the authors of Shrishti et al.
(2018). Fast loan approval for applicants was the main objective of this program. Logistic Regression,
Decision Trees, and Random Forest were the three machine algorithm types they employed. The Random
Forest algorithm was found to have the highest accuracy of all the models after examining the data sets for
various models.

Author Ndayisenga, T., (2021) bestowed the study with financial banks to foretell the nature of debtors
byunfolding and assessing the efficacy of numerous models using data from a Bank in Kigali. Training and
test datasets were produced by splitting the dataset into the ratio of 70:30. Association was utilized to
determine the most efficient machine learning models for predicting defaulters for any bank. The two most
effective models for predicting loan nonpayers were discovered to be gradient boost (accuracy 80.40%) and
XGBoost, while decision tree, logistic regression, and random forest did not perform well.

3. METHODOLOGY

Loan eligibility approval systems in the banking sector can be beneficial for both customers and the bank.
Machine learning approaches can ease the whole process as they will be cost-effective, and efficient and will
save time. In this work, important characteristics that are necessary for predicting loan eligibility are discussed
(Reddy et al., 2022).The dataset used in this study was gathered from a public repository. Training and
testing datasets are created after pre-processing the data to increase its overall quality (Gupta et al., 2020).
Machine learning models are developed using training data, while test data are used to assess the model’s
performance (Nayak et al., 2022). Decision tree, random forest, support vector machine, K-nearest neighbor,
Naïve Bayes, logistic regression, and linear regression are deployed here and their effectiveness in predicting
loan eligibility is evaluated (Kadam et al., 2021).

4. IMPLEMENTATION

A common Python distribution used for data science and scientific computing, Anaconda comes with a
graphical user interface (GUI) named Anaconda Navigator
(https://docs.anaconda.com/free/navigator/index.html). Jupyter Notebook is only one of the tools and
environments that can be managed and launched with ease using Anaconda Navigator.Jupyter Notebook is a
potential implementation option (https://jupyter.org/). It is interactive and features a user-friendly
environment for creating machine learning models.

4.1 Dataset
The dataset for predicting loan eligibility is gathered from the Kaggle public repository. Thirteen (13)
characteristics and 614 records build up the dataset. Loan id, Gender of the applicant, marital status of the
applicant, dependents in the family, qualification, employment status, applicant’s income, co-applicant income,
loan amount and term, credit worthiness, property area, and status of loan are the features present in the dataset.
The table below shows all the features with their description.

Table 1: features and their description

Attributes Description
Loan_id A unique ID
Gender Male/Female
Married Married/unmarried
Education Graduate/Undergraduate
Self_employed Yes/No
Dependents Number of dependents
Applicant_income Applicant’s income

50
Global Journal of Modeling and Intelligent Computing (GJMIC) ISSN: 2767-1917, Vol. 3, No.1, January 2023

Coapplicant_income Co-applicants income

Loan_amount Loan amount
Loan_term Loan’s repayment period
Credit_history Creditworthiness
Property_area Urban/rural/semi-urban
Loan_status Yes/No

4.2 Data Pre-processing

For the models to perform well, the dataset needs to be pre-processed. In this work, we used four Pre-
processing techniques which are as follows.

4.2.1 Handling missing values:Our dataset has many missing values and to handle that we used two
functions namely bfill() and fill() to fill all the null values so that our dataset will work efficiently (Yuvarani
et al., 2023).
4.2.2 Label encoding:Our dataset had both categorical variables as well as numerical variables so to bring
uniformity we used labelencoder() to convert all the categorical variables into numerical format (Wang et al.,
2022).
4.2.3 Feature scaling:Feature scaling is a Pre-processing technique that is used to transform the dataset into a
certain range. There are two scaling methods, one is normalization and another one is standardization. We
have used standard scalar to standardize our dataset. It removes the mean and scales the variables to the unit
variance.
4.2.4 Linear discriminant Analysis:Linear discriminant Analysis is a Pre-processing step that reduces the
dimensionality of the dataset and transforms the dataset from 2 dimensional to 1 dimensional. It also reduces
the curse of dimensionality.
4.2.5 Feature correlation: Making data-driven judgements and seeing patterns requires an understanding of
the relationships between characteristics. Figure 2 displays the association between various features in the
dataset we used (https://www.kaggle.com/code/vinodkumargr/loan-prediction).

Figure 2: The correlation among all the attributes in the dataset

(https://www.kaggle.com/code/vinodkumargr/loan-prediction)

4.3 Machine Learning Models

We employed seven machine learning methods in this study: Linear Regression, Logistic Regression, Support
Vector Machine, Naïve Bayes, Decision Tree, K-nearest Neighbour, and Random Forest.

4.3.1 Linear Regression: Linear regression is one of the easiest and drastically used Machine Learning
algorithms. It is a manner of performing predictive assessment through the use of statistics. Linear regression
generates forecasts for continuous/real/numeric variables (Orji et al., 2022).
4.3.2 Logistic Regression: Logistic regression is one of the Machine Learning algorithms this is maximum
often hired within the Supervised Learning category. It is used to forecast the specific based variable and the
usage of a special set of impartial variables. Logistic regression is used to be expecting the output for a based
variable this is expressed. The final results have to consequently be a discrete or express value. It gives the
probabilistic values that lie between zero and 1 in preference to the perfect values between zero and 1. It may
be either True or False, zero or 1, or Yes or No. Logistic regression and linear regression are pretty similar,

51
Global Journal of Modeling and Intelligent Computing (GJMIC) ISSN: 2767-1917, Vol. 3, No.1, January 2023

apart from how they’re used. While logistic regression is used to deal with type problems, regression
problems are addressed with the aid of using linear regression
4.3.3 Support Vector Machine: The extreme vectors and points that help create the hyperplane are chosen
via SVM. The SVM approach is based on support vectors, which are utilized to represent these extreme
situations (Bansal et al., 2022).
4.3.4 Naïve Bayes:Naïve Bayes classifiers are a subset of classification algorithms based on the Bayes
theorem. Instead of being a single approach, it is a family of algorithms that all work by the same guiding
principle: that each pair of features is being classed stand-alone (Orji et al., 2022).
4.3.5 Decision Tree: Classification and regression troubles may be resolved with the usage of the supervised
studying approach called a choice tree, but this technique is often preferred. It is a tree-dependent classifier,
wherein every leaf node represents the category’s final results and internal nodes constitute the functions of a
dataset. The nodes in a choice tree are the Decision Node and Leaf Node. Decision nodes are used to make
choices and feature many branches, while Leaf nodes are the results of choices and do now no longer have
any extra branches (Bansal et al., 2022).
4.3.6 K-nearest Neighbour: The KNN technique locations the brand-new example inside the class that
resembles the modern-dayclasses the maximum, presuming that the brand-new case and the preceding
instances are comparable. After all of the preceding statistics have been recorded, a brand new statistics factor
is labelled the use of the KNN set of rules primarily based totally on similarity. This suggests that with the
KNN approach, new statistics can be reliably and unexpectedly labelled. The KNN method may be used for
regression even though category issues are wherein its far maximum normally applied (Bansal et al., 2022).
4.3.7 Random Forest: The random forest bases its prediction of the final outcome on the predictions that
garnered the most overall votes rather than using just one decision tree’s forecast. To increase the projected
accuracy of the input dataset, the Random Forest classifier averages the results from multiple decision trees
used on various subsets of the input dataset. Higher accuracy and overfitting are prevented by the larger
number of trees in the forest.

5. RESULT
In the banking sector, the loan has always been the greatest source of income for a bank but the loan approval
process consumes a lot of time if done manually. To automate this whole procedure, we accustomed seven
algorithms of machine learning. We assessed the efficacy of seven machine learning techniques in terms of
accuracy. Linear regression has the lowest accuracy of just 51%, while the Random Forest classifier
outscored all six other models, scoring the greatest accuracy of 90.71%. Table 2 summarizes the overall
machine learning models’ average accuracy.

Table 2: Accuracy of all the implemented models

Classifier Accuracy

Linear Regression 51.21%

KNN 54.09%
Logistic regression 74.83%
Decision Tree 78.62%
SVM 82.23%
Naïve Bayes 85.96%
Random Forest 90.71%

6. FUTURE WORK AND CONCLUSION

This work concluded that the dataset is incomplete and still lacks some feature vectors. No classifier was able
to perform better than 90.71% because the subspace of the enter area that we have been seeking to generalize
has unknown greater dimensions (Random Forest classifier). More characteristic vectors ought to be produced
inside the destiny if similar studies are completed to provide the dataset applied to this observationso that the
classifiers can construct higher expertise of the difficulty at hand. For improved accuracy and performance in
the future, our work aims to employ a machine learning model with some deep learning techniques like CNN.

52
Global Journal of Modeling and Intelligent Computing (GJMIC) ISSN: 2767-1917, Vol. 3, No.1, January 2023

The deployment of DL models may also reduce computational time as it requires fewer manual pre-
processing tasks.

REFERENCE

Al Mamun, M., Farjana, A., &Mamun, M. (2022): Predicting Bank Loan Eligibility Using Machine Learning
Models and Comparison Analysis. 7th North American International Conference on Industrial Engineering
and Operations Management, Orlando, Florida, USA, June 12-14.
Awodele, O., Alimi, S., Ogunyolu, O., Solanke, O., Iyawe, S., &Adegbie, F. (2022, November). Cascade of
Deep Neural Network And Support Vector Machine for Credit Risk Prediction. In 2022 5th Information
Technology for Education and Development (ITED) (pp. 1-8). IEEE.
Bansal, M., Goyal, A., &Choudhary, A. (2022). A comparative analysis of K-nearest neighbor, genetic, support
vector machine, decision tree, and long short term memory algorithms in machine learning. Decision
Analytics Journal, 3, 100071.
Gupta, A., Pant, V., Kumar, S., & Bansal, P. K. (2020, December). Bank Loan Prediction System using
Machine Learning. In 2020 9th International Conference System Modeling and Advancement in Research
Trends (SMART) (pp. 423-426). IEEE.
https://jupyter.org/ [Last accessed on 15-06-2023]
https://docs.anaconda.com/free/navigator/index.html [Last accessed on 10-05-2023]
https://www.kaggle.com/code/vinodkumargr/loan-prediction [Last accessed on 01-05-2023]
Kadam, A. S., Nikam, S. R., Aher, A. A., Shelke, G. V., &Chandgude, A. S. (2021). Prediction for loan
approval using machine learning algorithm. International Research Journal of Engineering and Technology
(IRJET), 8(04).
King, T., &Frishberg, I. (2001). Big Loans, Bigger Problems: A Report on the Sticker Shock of Student Loans.
Kumar, C. N., Keerthana, D., Kavitha, M., &Kalyani, M. (2022, June). Customer Loan Eligibility Prediction
Using Machine Learning Algorithms in Banking Sector. In 2022 7th International Conference on
Communication and Electronics Systems (ICCES) (pp. 1007-1012). IEEE.
Mohankumar, M., Amuthakkani, S., &Jeyamala, G. (2016). Comparative analysis of decision tree algorithms
for the prediction of eligibility of a man for availing bank loan. Age, 19, 60.
Nayak, D. S. K., Routray, S. P., Sahooo, S., Sahoo, S. K., & Swarnkar, T. (2022, August). A Comparative Study
using Next Generation Sequencing Data and Machine Learning Approach for Crohn's Disease (CD)
Identification. In 2022 International Conference on Machine Learning, Computer Systems and Security
(MLCSS) (pp. 17-21). IEEE.
Ndayisenga, T. (2021). Bank loan approval prediction using machine learning techniques (Doctoral
dissertation).
Orji, U. E., Ugwuishiwu, C. H., Nguemaleu, J. C., &Ugwuanyi, P. N. (2022, April). Machine Learning Models
for Predicting Bank Loan Eligibility. In 2022 IEEE Nigeria 4th International Conference on Disruptive
Technologies for Sustainable Development (NIGERCON) (pp. 1-5). IEEE.
Park, M. S., Son, H., Hyun, C., & Hwang, H. J. (2021). Explainability of machine learning models for
bankruptcy prediction. IEEE Access, 9, 124887-124899.
Pandey, N., Gupta, R., Uniyal, S., & Kumar, V. (2021). Loan approval prediction using machine learning
algorithms approach. International Journal of Innovative Research in Technology, 8(1), 898-902.
Reddy, C. S., Siddiq, A. S., &Jayapandian, N. (2022, June). Machine Learning based Loan Eligibility Prediction
using Random Forest Model. In 2022 7th International Conference on Communication and Electronics
Systems (ICCES) (pp. 1073-1079). IEEE.
Samreen, A., & Zaidi, F. B. (2012). Design and development of credit scoring model for the commercial banks
of Pakistan: Forecasting creditworthiness of individual borrowers. International Journal of Business and
Social Science, 3(17).
Sheikh, M. A., Goel, A. K., & Kumar, T. (2020). An approach for prediction of loan approval using machine
learning algorithm. In 2020 International Conference on Electronics and Sustainable Communication
Systems (ICESC) (pp. 490-494). IEEE.
Singh, V., Yadav, A., Awasthi, R.,&Partheeban, G. N. (2021). Prediction of modernized loan approval system
based on machine learning approach. In 2021 International Conference on Intelligent Technologies
(CONIT) (pp. 1-4). IEEE.
Supriya, P., Pavani, M., Saisushma, N., Kumari, N. V., &Vikas, K. (2019). Loan prediction by using machine
learning models. International Journal of Engineering and Techniques, 5(2), 144-147.
Srivastava, S., GARG, A., SEHGAL, A., & KUMAR, A. (2018). Analysis and Comparison of Loan Sanction
Prediction Model using Python. International Journal of Computer Science Engineering and Information
Technology Research (IJCSEITR), 8, 1-8.

53
 Global Journal of Modeling and Intelligent Computing (GJMIC) ISSN: 2767-1917, Vol. 3, No.1, January 2023

Tejaswini, J., Kavya, T. M., Ramya, R. D. N., Triveni, P. S., &Maddumala, V. R. (2020). Accurate loan
approval prediction based on machine learning approach. Journal of Engineering Science, 11(4), 523-532.
Wang, C., Yang, N., Xu, W., Wang, J., Sun, J., & Chen, X. (2022, July). Research on a text data preprocessing
method suitable for clustering algorithm. In 2022 3rd International Conference on Information Science,
Parallel and Distributed Systems (ISPDS) (pp. 340-344). IEEE.
Yuvarani, P., Bharani, P., Dharun, B., & Dinesh, P. (2023, March). Time Series Forecasting of Ethereum Price
by FB-Prophet. In 2023 4th International Conference on Signal Processing and Communication (ICSPC)
(pp. 272-277). IEEE.

54

View publication stats