Based on the recycling technology of waste concrete, the preparation of regenerated aggregate and regenerated fine powder, and the use of them to replace quartz sand and cement, and the preparation of regenerated ultra-high performance concrete (R-UHPC) is a research hotspot in recent years. In this paper, a new machine learning method is proposed to predict the compressive strength performance of R-UHPC by using XGBoost algorithm. The compression performance of R-UHPC is predicted by using cement substitute material and aggregate substitute material index. More than 200 groups of literature data at home and abroad are trained, and the good fit is above 0.93, and the error between the predicted value of the model and the literature value is within 5%. The error between the model prediction value and the literature value is within 5%, indicating that the established XGBoost model has good prediction effect. By establishing the Python program of improved MMA matching, 21 groups of specimens were made, and the mechanical test study was carried out. The error between the predicted value of the model and the actual value of the test was within 5%. The results show that the established XGBoost model has good prediction effect