International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 09 | Sep 2024
p-ISSN: 2395-0072
www.irjet.net
CREDIT CARD SCORING ANALYSIS USING MACHINE LEARNING AND DEEP LEARNING Anmol K A, K D Sruthi Anmol K A , Msc. Computer science, St.Thomas(Autonomous)College, Thrissur 680001,Kerala , India K D Sruthi , Msc. Computer science, St.Thomas(Autonomous)College, Thrissur 680001,Kerala, India --------------------------------------------------------------------------***---------------------------------------------------------------------------online behavioral data has demonstrated superior Abstract - In today’s world, credit scores are essential to
performance compared to LSTM and traditional machine learning models. Our innovative end-to-end deep learning credit scoring framework incorporates both credit feature data and user behavioral data. The framework comprises a wide part and a deep part, enabling automatic learning from user data to enhance decision-making in credit granting.
determine credit worthiness for lending institutions, and they impact everything from getting a mortgage to renting an apartment. This thesis addresses key challenges in credit scorecard development, focusing on three main contributions. Firstly, it evaluates the performance of supervised classification techniques on imbalanced credit scoring datasets. Secondly, it explores the low-default portfolio problem, a severe form of class imbalance in credit scoring. Thirdly, it quantifies differences in classifier performance across various implementations of a real-world behavioral scoring dataset. Additionally, the thesis demonstrates the use of artificial data to overcome challenges associated with real-world data, while acknowledging the limitations of artificial data in evaluating classification performance.
1.1 Credit Scoring Process Credit scoring which is a conventional decision model and it is mainly focusing on risk approximation approach associated with credit products such as credit card, loans, etc. and is estimated based on applicant’s historical data which helps credit lenders in granting credit products.The following diagram illustrates the flow of data through the credit scoring process, from data collection to the final credit score output.
Keywords: Credit scoring, Machine learning, Deep learning, FE-Transformer, Feature selection.
1.INTRODUCTION Credit scoring models play a crucial role in the business landscape, providing a numerical assessment of an individual's creditworthiness based on diverse financial factors. Lenders and credit card companies heavily rely on these scores to make informed decisions on loan approvals or credit extensions. Typically ranging from 300 to 850, higher scores signify better creditworthiness. Scores above 700 are generally considered good, while those below 600 are seen as poor. Despite meticulous verification processes, there's no absolute assurance that credit cards are granted only to deserving candidates, emphasizing the ongoing importance of refining credit assessment strategies. Credit scoring serves as a conventional decision model, primarily focusing on risk assessment related to credit products like credit cards and loans.Financial institutions are increasingly embracing diverse risk assessment tools, including statistical analysis, to minimize potential risks. The utilization of deep learning algorithms, specifically transformers, based on
© 2024, IRJET
|
Impact Factor value: 8.315
Fig-1:Data Flow Diagram
|
ISO 9001:2008 Certified Journal
|
Page 124