Analyzing Customer Data and Predicting Non Performing Loan Score

Abstract

"A major risk to the sustainability and profitability of financial institutions are nonperforming loans, or NPLs. This project suggests incorporating a Non-Performing Loan Prediction System into an existing loan management system of a well-known financial institute in Sri Lanka. The project's main goal is to provide the financial institute with a predictive tool for determining the probability that borrowers will default on their loans when facilities are being granted. By implementing a machine-learning based prediction model, the institute hopes to improve loan approval, decision-making procedures and strengthen its risk management strategies. The identified problem is that the financial institute does not have a specific model for predicting non-performing loans. Officers do not now have a systematic method to determine the probability of a customer's loan would become non-performing over time. This gap results in the institute's incapability to predict and minimize possible credit risks within the institute. The proposed solution involves implementing a non-performing loan prediction system and integrate it with the existing loan management system. At the point of loan approval, the system aims to provide a Non-Performing Loan score for every customer using machine learning algorithms. The research will make use of supervised machine learning algorithms with a variety of datasets that include customer profiles, historical loan information, and other relevant features. Algorithms such as Random Forest and Logistic Regression will be trained and evaluated to identify the most accurate model for predicting Non-Performing Loans. In conclusion, the financial institute will receive many significant advantages from the implementation of the NPL Prediction System. This can result in lesser non-performing loans (NPLs), better portfolio performance overall, and a stronger financial position for the organization."