Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR292

Loan Approval Prediction System Using Machina

Learning
P V V S V Prasad 1 ; P.V Nageswara Rao2
1, 2Departement of Computer Science and Engineering, GITAM School of Technology,
GITAM (Deemed to be University) Visakhapatnam, Andhra Pradesh, India

Abstract:- With the growth of the banking sector, the For banks, lending is the main area of work lending,
identification of reliable borrowers that must maintain money changing, taking deposits from customers, and
the natural core income and asset accumulation becomes earning profits by making interest from customers' loans.
a key issue. Despite all security measures, reliability of The goal here is the funding of creditworthy borrowers, but
customers remains an unclear question. To tackle this standard procedures through lending banks fail to provide
barrier, banking management which is directed towards protection for selected borrowers. To solve this problem, we
customer loan repayment consistency is required. Credit designed a credit finding algorithm with the help of Python
approval is a significant aspect of economy since it and began to get credit approvals from banks. This project
determines the allocation of credit-linked funds. Today, does an eligibility analysis of a client by a process of
machine learning is known for its power to automate and evaluating different parameters including marital status,
scale up the processing of application for loans. income, and expenditure to determine whether they meet the
capacity to repay the loan or not. We use the creates trained
This project will begin with data collection which data sets as a model that we directly then test the data to
will consist of data on historical (regarding the past) loan output a result of "yes" or "no" only to determine if a
applications and the borrower profiles. The dataset has customer or not. can repay the loan. It goes without saying
features of the credit score, income, previous work that This loan approval process is good business sense.
experience, debt-to-income ratio, and loan repayment
record. This way, the models learn through the strengths Automating approvement reduces the use of electricity,
to find good features and the reasons for accepting a upgrades the quality, and makes the customer satisfaction
loan. They are the experts in these areas and can forecast better. It can do so through machine learning and hence can
the potential patterns and connections of the data. do more precise and consistent loan decisions. This
Within the scope of this work, the supervised algorithms approach facilitates just and accurate loan approvals. ML
used are logistic regression, decision trees, random models are a solution to fair analysis and cut off human
forests, and support vector machines. These algorithms biases and errors in manual data analysis.
are applied to the dataset available to often produce
results like binary classification and regression. The II. METHODOLOGIES AND TECHNIQUES
adoption of machine learning among financial
institutions is intended for a faster processing of loans The supervised learning methods are employed to train
which is their benefit. What is credit scoring, it is a tool the prediction model in this loan approval ML task. We
which automate manual loan application review thereby research and study different machine learning algorithms
increases efficiency. The machine-learning algorithms such as logistic regression, decision trees, random forests,
that analyse applications for loans could cut down on the and support vector machines. Feature selection methods and
possibilities of human biases and mistakes which are an data preparation techniques are used to improve model
inherent part of the process. Also, ML uses the model to performance and learning. Undefined.
recognize borrowers who may default and subsequently
lower the likelihood of default. Part of the task involve The execution of the mentioned program may be
utilizing historical credit market data and implementing beneficial for banks and borrowers at the same time. The
ML algorithms to develop a highly accurate and reliable automated loan approval process simplifies decision-
loan approving system based on trained-data, random making, saves manual work and improves efficiency in the
forests, the stream of loans and reliable clients. financial institutions’ operations. ML model facilitates risk
management through precise identification of high-risk
Keywords:- Construction of Data, Creation of Correlation, customers, leading to fewer defaults and losses. Borrowers
Banker Loans, Customer Safety. appreciate a fair and clear loan approval system, in which
many factors are taken into consideration and credit scores
are not the key consideration in a holistic credit evaluation.
This paves the way for the ones with little or no credit
I. INTRODUCTION history or non-financial characters. ML techniques will be
used as a basis for developing a loan approval mechanism

IJISRT24APR292 www.ijisrt.com 278

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR292

that is more precise and operationally efficient for both support vector machine (SVM) algorithms combining so
lenders and borrowers. Provides predictive accuracy.

In this sphere we face data biases from historical cases E. Forecast Default for Networked- Guaranteed Loans
showing the gap in practices while the machine learning Dawei Cheng, Zibin Niu, Yi Tu, and Liqing Zhang
models are accurate in the quantitative domain but may be address the challenge of predicting the risk of guaranteed
less context sensitive and very complex. On top of being loans in a network by applying an imbalanced approach to
difficult to comply with, there exists the risk of over– the spread of a network with its neighbour a well-weighted
reliance on models. (pwkNN) to be used to forecast cooperative risk in their
proposed model) and algorithms are included.
To cope with these issues a fairness must be ensured,
and a coordinated and comprehensive strategy should be F. Personalized Credit Rating Using Artificial Intelligence
used that includes both human qualified experts. The loan Technology for Federal Student Loans
approval process involves a detailed consideration. It Jian Hu describes the development of a loan
considers contextual factors and complies with regulatory management system designed for federal student loans using
guidelines. The objective in mind is making choices Artificial Neural Network (ANN) technology. Leveraging
between risks and opportunities wisely. the adaptive capabilities of ANN, proposed back
propagation neural network reveals effective credit rating
III. LITERATURE SURVEY evaluation for college students despite the presence of credit
history.
A. An Approach Using Machine Learning Algorithms for
Loan Approval Prediction IV. DATA ANALYSIS
The study by Mohammad Ahmed Sheikh, Amit Kumar
Goel and Tapas Kumar highlights the importance of The course starts with a thorough examination of a
predicting defaults of banking systems to reduce non- historical credit data that contains attributes of applicants as
performing assets (NPAs) and increase profitability. The well as their credit outcomes. The data set has been carefully
paper compares different forecasting methods, with a crafted and cleaned for missing values, redundancies, and
particular focus on Logistic Regression models. class imbalance issues.

Through the study it is concluded that the addition of Exploratory data analysis methods are used to study
individual characteristics to traditional checking account the distribution of loan permissions among different
information improves forecasting accuracy. This highlights demographic groups, types of loans, and risk types, and
the importance of considering a range of consumer factors feature engineering techniques to get additional knowledge
for loan approval decisions. and the accuracy of ML models rises.

B. Default Forecasting of Loans Through Data Mining The study starts with a comprehensive study of
Bhumi Patel, Harshal Patil, Jovita Hembram, and Sri historical credit data, which includes various components on
Jaswal propose the use of data mining algorithms to predict applicants’ financial background and credit outcomes The
loan defaults. By using datasets containing home loan data set is carefully crafted and pre-processed to improve its
applications, the paper aims to help banks make more standards of relevance, duplicates, and class imbalance.
rational decisions.
With exploratory data analysis we determine how the
The study highlights the need for accurate accounting loan approvals are distributed through different demographic
to minimize losses and increase loan approvals. groups, loan types and risk types and by applying feature
engineering techniques we gain useful information and
C. Forecasting Loan Position in Commercial Banks by improve our ML models performance.
Machine Learning Classifier
G. Arutjothi, Dr. C.S. Senthamarai will focus on V. MODEL DEVELOPMENT
building credit scoring models for commercial banks using
machine learning techniques. The paper recommends a The fundamental part of the research is the
machine learning classification-based approach, where Min- development of ML algorithms and the estimation of their
Max normalization and K-Nearest Neighbour (K-NN) accuracy in classification of loan applicants. The algorithms
algorithm are combined the goal is to predict the loan used, such as logistic regression, decision trees, random
condition more accurately, thus helping in making credit forest, and gradient boosting, are many.
decisions.
With the use of boosting the machines and neural
D. Bank Loan Overpayment Prediction Based on LSTM- networks are getting trained and improved with cross-
SVM validation method. Hyperparameter tuning and ensemble
Xin Li, Xianzhong Long, Guozi Sun, Geng Yang, and learning approaches are included to increase the model
Huakang Li propose a new method for bank loan default accuracy, stability, and generalization capability.
prediction using long-term short-term memory (LSTM) and

IJISRT24APR292 www.ijisrt.com 279

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR292

Similarly, model interpretability is critical as a variable respectively. The decision tree classification model achieved
in this regard as it allows the processes of a loan to be an accuracy of above 90%, with precision, recall and F1
understood and the procedure to remain transparent. scores of 0.91, 0.90 and 0.90 respectively.

F1 scores of 91%, 90%, and 90% for each dataset.

Random forest and the k-nearest neighbour classification
algorithms had precision, recall, and F1 scores of 92% and
82% respectively which did not significantly vary from
those of performance levels of other classifiers. These
results provide financial institutions which are seeking to
implement loan approval prediction algorithms an effective
framework for decision making.

Table 2 Results
Metric Value
Logistic Regression Accuracy 72%
Recall 86%
Precision 70%
Fig 1: Workflow of Proposed System
F1 Score 65%
VI. IMPLEMENTATION AND RESULT Decision Tree Accuracy 90%

 PyTorch: Recall 95%

A deep learning framework that consists of developing Precision 95%
and training neural network models.
F1 Score 91%
 NumPy:
Random Forest Accuracy 92%
Essential for doing numerical computations and
handling arrays. Recall 95%

 Pandas: Precision 95%

Enables data management and analysis, including the K-NN Accuracy 92%
handling of large data sets.
Recall 95%
 Matplotlib:
Precision 95%
Introduces data visualization and performance metrics
of the model.
VII. CONCLUSION
 Scikit-Learn:
Finally, this study compares classification models
Include evaluation metrics and data pre-processing
applicable for loan approval prediction and develops
while undertaking machine learning operations.
implications that relate to the models' performance and
suitability. The research results show that each model
Table 1 Implementation
possesses unique strengths and weaknesses, but it is
NumPy Numerical computations andarray revealed by the data that both decision trees classification
manipulation and forest random forests classification are predictors of
loan outcomes that could be most reliable and accurate.
Pandas Data manipulation andanalysis Employing these models, financial institutions can
streamline the decision-making processes, automate credit
Matplotlib Data visualization scoring, and conduct risk assessments, enhancing their
Scikit-Learn Machine learning tools andevaluation customer satisfaction and operational profits.
metrics

 RESULTS FUTURE RESEARCH

The distribution mentioned for each model serves as a
reminder of its ability to accurately classify loan approvals. The future studies should concentrate on sophisticated
The logistic model demonstrated the accuracy of 72%. Its machine learning techniques like neural networks for loan
accuracy, recall, and F1-score were 73%, 72%, and 72% approval predictions. On the other hand, the introduction of

IJISRT24APR292 www.ijisrt.com 280

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR292

new data sources and advanced feature engineering

techniques can lead to a greater precision of models. In
addition to that, research efforts can better be aimed at
creating machine learning models that are interpretable to
increase the transparency and confidence in the loan
approval decision-making processes.

REFERENCES

[1]. K. Hanumantha Rao, G. Srinivas, A. Damodhar, M.

Vikas Krishna. "Implementation of Anomaly Detection
Technique Using Machine Learning Algorithms."
International Journal of Computer Science and
Telecommunications, Vol. 2, Issue 3, 2011.
[2]. S.S. Keerthi, E.G. Gilbert. "Convergence of a
Generalized SMO Algorithm for SVM Classifier
Design." Machine Learning, Springer, Vol. 4, Issue 1,
pp. 351-360, 2002.
[3]. Andy Liaw, Matthew Wiener. "Classification and
Regression by Random Forest." Machine Learning,
Vol. 2, Issue 3, pp. 9-22, 2002.
[4]. Ekta Gandotra, Divya Bansal, Sanjeev Sofat. "Malware
Analysis and Classification: A Survey." Journal of
Information Security, Vol. 05, Issue 02, pp. 56-64,
2014.
[5]. Rattle Data Mining Tool. [Online] Available:
http://rattle.togaware.com/rattle-download.html.
[6]. Aafer Y., Du W., Yin H. "Droid APIMiner: Mining
API-Level Features for Robust Malware Detection in
Android." Security and Privacy in Communication
Networks, Springer, pp. 86-103, 2013.
[7]. J. R. Quinlan. "Induction of Decision Tree." Machine
Learning, Vol. 1, No. 1, pp. 81-106.

IJISRT24APR292 www.ijisrt.com 281