Several rating agencies such as Standard & Poor's (S&P), Moody's and Fitch Ratings have evaluated firms’ credit rating. Since lots of fees are required by the agencies and sometimes the timely default risk of the firms is not reflected, it can be helpful for stakeholders if the credit ratings can be predicted before the agencies publish them. However, it is not easy to make an accurate prediction of credit rating since it covers a variety of range. Therefore, this study proposes two double ensemble approaches, 1) bagging-boosting and 2) boosting-bagging, to improve the prediction accuracy. To that end, we first conducted feature selection, using Chi-Square and Gain-Ratio attribute evaluators, with 3 classification algorithms (i.e., decision tree (DT), artificial neural network (ANN), and Naïve Bayesian (NB)) to select relevant features and a base classifier of ensemble models. And then, we integrated bagging and boosting methods by applying boosting method to bagging method (bagging-boosting), and bagging method to boosting method (boosting-bagging). Finally, we compared the prediction accuracy of our proposed model to benchmark models. The experimental results showed that our proposed models outperformed the benchmark models.