To predict the employability of master degree candidates, a corresponding employability model based on the optimized XGBoost algorithm is constructed. First, 16 factors affecting employment are selected based on the training process of master degree candidates, such as the weighted grades of degree courses, the approval of patent applications, the participation in scientific research projects, the publication of scientific research papers, the acquisition of skill certificates and excellent theses. Secondly, the parameters of XGBoost algorithm are optimized by combining the artificial bee colony algorithm and the five-fold cross-validation method. Based on the optimized XGBoost algorithm, the training data and employment data of master degree candidates are analyzed. Through empirical research, it shows that the proposed model outperforms other models in terms of accuracy rate/misjudgment rate. According to the feature importance ranking output by the prediction model, scientific decision support can be provided for training institutions of master degree candidates.