Risk early-warning is a very important task in earthquake emergency response. It is a complex prediction problem, often with inconsistent and even conflicting factors. With the integration of the matter-element theory and entropy method, we propose a new approach that integrates the early-warning objects, risk attributes and their value range into a matter-element system, to research risk early-warning system’s extension nature. Through the application of extension theory, the weight coefficients of early-warning attributes are determined by using an entropy weight method. The relational degree of the early-warning attributes and comprehensive risk early-warning degree are then calculated. As a case study, the approach is applied to the landslide risk early-warning. The results demonstrate the rationality, feasibility, and impersonality of our approach.