As one of emerging ceramic materials, high-entropy ceramics become a research hotspot in the field of ceramics. However, their compositional design has some challenges for component design based on experimentation and “trial and error”. In recent years, the combination of machine learning and experiments can provide an effective method to solve this problem. In this paper, four machine learning models were established, the best-performing gradient-boosting decision tree model (R2=0.92) through training and evaluation was selected for prediction. A single-phase (Ti0.2V0.2Zr0.2Nb0.2Hf0.2)N high-entropy nitride ceramic was then synthesized based on the predication by the model. This effective approach can provide some ideas for the design of high-entropy nitride ceramic and discover new systems.