Korea Asset Management Corporation (KAMCO) announced on the 17th that a research paper predicting the repayment ability of vulnerable debtors in Korea has been published in the international academic journal Computational Economics.
This journal, first published in 1988, is issued six times a year focusing on solving economic problems using computer science. It is indexed in SSCI (Social Sciences Citation Index).
Through this sole research, KAMCO is recognized for effectively predicting the final repayment ability of vulnerable debtors using the stacking algorithm, a machine learning methodology, and for providing implications for efficient non-performing loan management based on debtor characteristics.
The paper identified the relationship between creditor and debtor characteristics such as lending institution, loan size, and debtor age, and repayment ability. It confirmed that the final recovery outcome of non-performing loans can be predicted with a high accuracy of 87.7% using the stacking algorithm model. The stacking algorithm is a methodology that retrains results estimated by multiple machine learning models to derive a final outcome, offering the advantage of effectively compensating for the inherent issues of individual models.
Through this research, KAMCO expects to be able to identify in advance the types of loans and debtors with high recovery potential, thereby contributing to the enhancement of capabilities such as providing customized debt adjustment programs for debtors and selecting recoverable loans.
Kwon Nam-ju, President of KAMCO, stated, "As the ratio of non-performing loans across the financial sector is rising due to the economic downturn, proactive risk management is being emphasized. We will continue to monitor the financial market situation and actively seek ways to support the swift recovery of vulnerable groups."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
