With the development of the Internet, online credit has grown rapidly. At the same time, the low threshold for online credit has caused some problems, such as illegal fund raising and telecommunication fraud. In order to solve the above problems, machine learning algorithms have been used as an effective way to predict the credit defaults. In this paper, we use real data from Internet credit platforms as the sample set. We choose Random Forest (RF) algorithm and XGBoost algorithm to build separate default prediction models, using Blending method to fuse these two single models to construct a well-performing credit default prediction model. After evaluating the prediction results, the fused models outperformed the single models in default prediction.