This paper employs a large assortment of machine learning methods to dissect the cross-sectional variation in corporate bond returns in China. Using a rich set of macroeconomic and bond and issuer characteristic variables, our findings provide strong support for the outstanding performance of machine learning techniques in comparison with traditional linear models. The improvement brought by machine learning methods is statistically and economically meaningful, evidenced by the abnormal returns to bond portfolios based on predictions from machine learning.