<p>To avoid the impact of rockfall on railroad safety, it is crucial to accurately predict the risk level of rockfall hazard in advance for the safe construction, operation, and maintenance of railroads. We take the western mountainous railroad tunnel portals as the research background and construct a western mountainous railroad tunnel portals rockfall hazard risk evaluation index system, including 15 core indexes and 6 alternative indexes. The unascertained measure theory (UMT) is chosen to evaluate the risk of rockfall hazard in tunnel portals. The improved G1 method and entropy weight method (EWM) are introduced to calculate the risk of rockfall hazard based on the confidence level criterion. Furthermore, the criterion is used to recognize the risk level of a rockfall hazard. Ten typical tunnel portals of a railroad in the west are selected to validate engineering examples, and the evaluation results of the improved G1–EWM–UMT model are compared with those of the TOPSIS model [the technique for order preference by similarity to the ideal solution (TOPSIS)] and the actual rockfall situation in the project. The comparison results show that improved G1–EWM–UMT model has 90% accuracy of rockfall risk evaluation results with actual rockfall grade, which is 30% and 20% higher than the accuracy of TOPSIS model and extension cloud model, respectively. This study has certain application value and engineering significance, which provides a theoretical basis and technical reference to design the rockfall protection structure of western mountainous tunnels such as the Sichuan–Tibet railway under similar engineering geological conditions and proves the operability and applicability of the evaluation model.</p>