In this paper, an accurate prediction model, which screened from 8 machine learning-based models (LR, ANN, DT, SVC, RF, AdaBoost, GBDT, XGBoost), is established for identifying the failure mode of flat slabs. A database contains 610 experimental data is collected. The hyper-parameters are determined through grid search method with 10-fold cross validation, and precision, recall, F1 score and accuracy are utilized for appraising the prediction of each model. After comparison with other 7 machine learning-based models and 3 empirical models, XGBoost is selected as the best model, in which the precision, recall, F1 score and accuracy of which are 97.30%, 94.74%, 96.00% and 99.02%, respectively. The prediction of XGBoost is explained by SHAP, the results including global and individual interpretations and the feature dependency relationship between input variables. According to these results, the relationship between failure mode of flat slabs and influence factors is exhibited through another perspective.