A1088
Title: Corporate bond return prediction: An ensemble learning approach
Authors: Albert Zhao - Nankai University (China) [presenting]
Shan Jiang - Nankai University (China)
Abstract: Using predictors from corporate bond, Treasury, and stock markets, we find that corporate bond returns are predictable based on an ensemble learning approach, "Stacking". We show that while closely related to traditional combination forecast methods, the Stacking method generates higher statistical and economic gain across bond ratings and maturities by introducing new features into combination forecasts. The method is also insensitive to the problem of high dimensionality in that we achieve the best result when using all predictors jointly. While the overall performances of different Stacking models are satisfactory, simpler Stacking models perform better.
