To accurately evaluate the roof water inrush risk of deep roadways and make up for the low credibility of traditional evaluation methods caused by index redundancy and subjective weighting, a comprehensive calculation method coupling with principal component analysis (PCA) and entropy method was proposed. PCA was used to reduce the dimension of multi-source indexes to eliminate information overlap. After key principal components were extracted, the weight of each principal component was quantified. The entropy method was introduced to achieve the objective calculation of comprehensive evaluation indexes. Combined with the sample input datasets of each index, the evaluation model of coal seam roof water inrush was constructed with the logical framework of "principal component extraction-comprehensive index calculation-model comparison and verification-dynamic prevention and control". The results show that the fault water conductivity coefficient is the main controlling factor affecting the roof water inrush of the working face, and the permeability coefficient is the second most important. Based on the high fitting characteristics of the comprehensive index calculated by the PCA-entropy method and the on-site water seepage data, five categories of water-rich risk levels have been divided. It is determined that the area 518-678 m away from the open-off cut of the 1315 working face is in the extremely high-risk area, and the area 0-425 m away is in the high-risk area. According to the characteristics of roof water inrush in different areas, the technical path of "partition-based prevention and control-three-dimensional interception-drainage optimization" was proposed. The research results improve the classification and identification system of roof water inrush risk levels, and provide a new method for the safe exploitation of deep resources.