Published Paper Details:

The banking system claims that everyone's primary source of income comes from loans. A bank's primary source of revenue is, therefore, loans. One of the greatest threats to a bank's viability and profitability in today's cutthroat market is the difficulty of accurately assessing the risk associated with a loan application. Many individuals visit different banks every day to seek for loans. Every applicant does not get approval. Most banks utilize their risk assessment and credit scoring systems to determine whether or not to approve a customer's loan application. Getting a loan approved is a tedious process for bank employees. A more efficient, accurate, and quick loan approval procedure is possible with the use of modern technologies like machine learning models. Predicting whether or not a borrower will get a loan is an important yet long-standing problem for the financial services sector. Historically, banks and other lenders relied on manual processes and subjective criteria to evaluate loan applications, which often led to inconsistent decisions and increased risk of loan defaults. With the rise of ML algorithm, there is now an opportunity to create more accurate and reliable predictive models that can help financial institutions make better lending decisions. This research uses ML techniques to identify trends in a shared dataset of loan-eligible individuals. The dataset is prepared by performing exploratory data analysis and data balancing. The explored algorithms include Cat Boost, Gradient boosting and XGBoost. Preprocessing data, selecting features, balancing data, training and testing, classification, and comparing performance using accuracy and precision as classification metrics are the main goals of the project. Recall and f1-score. Following Gradient Boost Classifier (86.39%) and XGBoost (84.61%), the results reveal that Cat Boost Classifier had the greatest accuracy at 88.16%. The findings show that ML algorithms have the ability to make loan approvals better and decrease the likelihood of defaults.