A Comparison of Neural Networks and Classical Discriminant Analysis in Anticipating Default among High-yield Bonds

摘要

As mentioned earlier, the last decade has seen a rapid growth of the high-yield bond market. Along with this increasing use of high-yield bonds to raise capital, there has also been a rise in the default rates. This has influenced a lot of issues like the degree of risk in bond investment, trustworthiness of the bond-ratings, and protection for investors holding securities. Even though certain areas like the link between bankruptcy and bond ratings has been extensively researched, there is a scarcity of studies relating to bond defaults. This study addresses the topic of bond defaults using two models: one using the classical multivariate discriminant analysis and the other using neural networks. Such models on bond defaults can be of immense practical value to investors of high-yield securities. Investors can use such models to preserve their investment before the bond goes into default. From a theoretical perspective, models such as the ones we developed using different tools can help researchers gain valuable insight into the nature of high-yield securities and the factors which affect their quality.