With the continuous progress of civil engineering technology, concrete has become the most widely used material in the civil engineering industry. Its concrete pressure resistance is the core indicator for measuring concrete performance. With the development of big data and artificial intelligence, machine learning algorithms are playing a significant role in concrete strength prediction. at the same time, precise predicting concrete compressive strength has important practical significance. Based on the XGBoost (Extreme Gradient Boosting) algorithm, this study has established a non-linear prediction model of concrete compressive strength. And by collecting and analyzing the setting data of 880 concrete groups, the differences in the predictive model of the four machine learning algorithms of XGBoost, Lightgbm, NGBOOST and Catboost were compared. The research results shown that the performance of models based on the XGBoost algorithm in training and test sets was better than that the other three models. The model based on the XGBoost algorithm has high predictive accuracy and good generalization capabilities. R², MAE, and RMSE based on the XGBoost algorithm models were 0.9873, 1.1707, and 1.8657, respectively. R², MAE and RMSE based on the XGBoost algorithm models were 0.9141, 3.2671, and 4.8232, respectively. The R², MAE and RMSEs obtained based on their data verification were 0.9655, 1.8952, and 2.6778, respectively. The model based on the XGBoost algorithm was excellent for predicting concrete compressive strength. A model based on the XGBoost algorithm was a reliable concrete compressive strength prediction model.