<p>Different modes of disasters will occur in the process of tunnel construction in thin-layered rock mass. The key to tunnel safety control is to analyze the main control factors affecting the failure mode of thin-layered rock mass tunnel and predict the potential failure mode. The failure characteristics of thin-layered rock mass were analyzed and classified based on 22 tunnel projects. In addition, the influence factors of different failure modes are quantitatively analyzed. Taking the influencing factors as the prediction index, combined with the actual case, the initial database is constructed. The Fisher-Freeman-Halton exact test method was used to calculate the main control factors of the failure mode. The initial database samples were pretreated by augmentation and CRITIC method weighting, and the failure mode prediction model of thin-layered rock mass tunnel was established by random forest. The evaluation metrics and validation index of the model is tested by practical engineering case, and the results are compared with RF, AdaBoost and SVM algorithms. The findings show that the failure modes of thin-layered rock mass tunnel can be divided into four types, including squeeze bending failure, buckling bending failure, along-layer slip failure, and falling block failure. The angle between the maximum stress and the rock strata and the degree of joint development are the main control factors affecting the failure mode. The accuracy, precision, recall and F1-Score of the CRITIC-RF are 81.8%, 81.7%, 80.2%, 0.809, and higher than AdaBoost and SVM. The Oob-Score of the CRITIC-RF is 0.824 and higher than RF. The reliability of the prediction results of the prediction model is ensured. The research results can provide the basis and reference for the failure mode prediction and active control of thin-layered rock mass tunnel. The research results can provide scientific basis for the prediction and precise control of engineering disasters in thin-layered rock masses.</p>