ARTIFICIAL INTELLIGENCE AND THE REFORM OF CREDIT RISK MANAGEMENT IN THE THAI CREDIT UNION SYSTEM : TRENDS, OPPORTUNITIES AND CHALLENGES

Abstract

This article aims to compile and synthesize knowledge related to enhancing credit approval processes and financial risk management through artificial intelligence (AI) and machine learning. It focuses on addressing the credit approval risk challenges faced by credit unions, financial institutions that currently do not utilize credit analysis systems. However, banks are increasingly using AI to analyze risk. This article covers the transition from traditional statistical models to advanced algorithms such as XGBoost and LightGBM, combined with alternative data analysis including digital behavior, utility payment history, and psychological data, to generate highly accurate credit scores and reduce non-performing loans (NPLs).

The article emphasizes explainable AI (XAI) techniques, such as SHAP and LIME, to overcome the black-box nature of AI, promoting transparency and reducing bias in accordance with good governance principles. It also considers data and the Personal Data Protection Act (PDPA), and presents a risk management framework for credit unions based on COSO ERM 2017 standards, the development of an early warning system, and guidelines for overseeing AI-related risks, in line with the Bank of Thailand's policy, to prepare for becoming a data-driven and sustainable organization.