Seasonal Analysis and Risk Management Strategies for Credit Guarantee Funds: A Case Study from Republic of Korea

This study investigates the prediction of small and medium-sized enterprise (SME) default rates in Republic of Korea by comparing the performance of three prominent time-series forecasting models: ARIMA, SARIMA, and Prophet. The research utilizes a comprehensive dataset provided by the Korea Credit Guarantee Fund (KODIT), which covers regional and monthly default rates from January 2012 to December 2023, spanning 12 years. By focusing on Republic of Korea’s 17 major cities, the study aims to identify regional and seasonal patterns in default rates, highlighting the critical role that regional economic conditions and seasonality play in risk management. The proposed methodology includes an exploratory analysis of default rate trends and seasonal patterns, followed by a comparative evaluation of ARIMA, SARIMA, and Prophet models. ARIMA serves as a baseline model for capturing non-seasonal trends, while SARIMA incorporates seasonal components to handle recurring patterns. Prophet is uniquely suited for dynamic datasets, offering the ability to include external factors such as holidays or economic shocks. This work distinguishes itself from others by combining these three models to provide a comprehensive approach to regional and seasonal default risk forecasting, offering insights specific to Republic of Korea’s economic landscape. Each model is evaluated based on its ability to capture trends, seasonality, and irregularities in the data. The ARIMA model shows strong performance in stable economic environments, while SARIMA proves effective in modeling seasonal patterns. The Prophet model, however, demonstrates superior flexibility in handling irregular trends and external events, making it the most accurate model for predicting default rates across varied economic regions. The study concludes that Prophet’s adaptability to irregularities and external factors positions it as the most suitable model for dynamic economic conditions. These findings emphasize the importance of region-specific and seasonal factors in tailoring risk forecasting models. Future research will validate these predictions by comparing forecasted default rates with actual data from 2024, providing actionable insights into the long-term effectiveness of the proposed methods. This comparison aims to refine the models further, ensuring robust financial stability and enhanced SME support strategies for institutions like KODIT.