Abstract: In order to scientifically and reasonably evaluate the risks of coal mine roof disasters, effectively prevent and reduce roof accidents during the deep coal mining process, a risk evaluation model based on entropy weight modified AHP-TOPSIS is constructed. This model establishes an evaluation system consisting of 20 items of second-level indicators from four dimensions like geological environment, technology, management, and behavior, the AHP weights are conducted to be modified through entropy weight method, the subjective limitations of single weighting method are effectively overcame; A four-level risk judgement criterion is established based on the critical value of indicator. The Monte Carlo simulation results show that the new model has relatively strong robustness to weight perturbations of ±10%, the probability of its level Ⅱ risk judgement reaches 96.7%, which is 16.6 percentage points higher than that of the traditional AHP method (80.1%). The actual case verification indicates that the model prediction result (Level Ⅱ moderate risk) are highly consistent with the actual case. Besides that, the sensitivity analysis identifies geological structures and empty roof operations as key sensitive indicators. This model provides a reliable quantitative decision-making tool for roof disaster prevention and control by combination weighting and stability optimization.