In order to study the corporate bond market and deal with the impact of exchange rate fluctuations on the corporate bond market, we propose to establish two neural network models of LSTM and GRU, first demonstrating that they have a good fit and prediction effect on the research of corporate bond market returns. Using the exchange rate data of RMB against the US dollar as one of the input variables of the neural network, we verify the impact of the RMB exchange rate on the corporate bond market returns and compare the prediction effects of the two models. The results show that after adding the exchange rate indicator, the two models can not only capture the yield trend, but also the value is more accurate and reliable. Empirical studies show that the RMB exchange rate is a non-negligible factor in the study of China's corporate bond market, and the results of the GRU model are more accurate.