Abstract: The problem of concrete shrinkage and cracking seriously endangers the structural safety and normal use of building engineering, and the accurate and rapid prediction of early crack resistance of concrete has become the research focus. In this paper, random forest combined with support vector machine algorithm (RF-SVM) was introduced into the study of early-age cracking resistance of concrete. Taking a project as an example, an index system of early-age cracking resistance of concrete was established by selecting 12 influencing factors based on material and mix ratio, in this paper, the random forest regression algorithm based on importance ranking was used to extract the features of the impact factors, select the optimal feature variable set, and achieve the goal of dimension reduction,at the same time, the factors that should be paid more attention to in the actual project were clarified. Then the parameters of the SVM model were optimized by the method of 10-fold cross-validation, and the selected samples were trained and predicted by the SVM model, and the predicted results were output, and compared it with the SVM model and the artificial neural network model without feature selection, the results showed that the prediction result of RF-SVM was the closest to the measured value and the model had the highest precision. The RF-SVM prediction model proposed in this paper could provide an effective method for rapid prediction of early crack resistance of concrete.