In order to reduce the harm caused by roadway roof fall, improve the efficiency of coal mining, and predict the risk of roadway roof fall timely and accurately, this paper proposes a risk assessment model of roadway roof fall based on grey correlation degree analysis and decision tree algorithm on the basis of 12 evaluation indexes. Firstly, the correlation degree between each index and the risk level is calculated based on grey relational analysis (GRA), and the evaluation indexes are screened according to the calculation results, and the strongly correlated indexes are selected. Then the machine learning algorithm of decision tree (DT) is used to establish the evaluation model of tunnel roof fall. In order to verify the validity of the model, the transport lane and return air lane of 122907 working face in Enhong Coal Mine are selected as cases to carry out risk assessment with the model, and the evaluation results are compared with the evaluation results of a single decision tree (DT) model. The gray correlation degree analysis shows that the correlation degree of 12 indicators of surrounding disturbance is smaller than that of the other 11 indicators, and the indicators are identified as weak correlation and removed. The evaluation results show that the evaluation results of this method are in line with the actual conditions, and the accuracy of the evaluation results of the single decision tree model is only 67%, and the evaluation results of the decision tree-grey relational degree model are better than the decision tree model. The grey correlation degree decision tree model effectively solves the interference problem of weak correlation indicators, and significantly improves the assessment accuracy of roadway roof fall risk. This method can provide a new technical path for mine roadway safety evaluation, and has the value of engineering popularization and application.