Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4946471 | Knowledge-Based Systems | 2016 | 8 Pages |
Abstract
Aiming to improve the KMV model as well as estimate the price discount on non-tradable shares, this paper proposes the methods of particle swarm optimization and maximum likelihood estimation to develop the PSO-KMV model. We find evidence that the PSO-KMV model significantly outperforms the KMV model and the market price of non-tradable shares estimated by the PSO-KMV model is reasonable. Besides, particle swarm optimization is more suitable than genetic algorithm in this study. Further, the method of fuzzy clustering is applied to the PSO-KMV model. Then the FC-PSO-KMV model obtains a more specific default point composed of four parts of liabilities instead of short-term debt and long-term debt, and thus performs slightly better than the PSO-KMV model. The empirical results demonstrate that the short-term debt of China, especially the short-term loan, is far riskier than that of the U.S. In summary, the hybrid KMV models provide more accurate predictions so as to effectively assess and manage credit risk.
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
Zhang Yaojie, Shi Benshan,