Abstract: Goaf safety evaluation is a typical multi -factor and non -compatible problem. In the paper, representative indexes were selected from three aspects of rock quality, environmental condition and goaf occurrence to built evaluation indexes system of goaf safety. Based on matter -element extension theory, matter-element extension model was established by matter-element construction and correlation solution. The origin model was improved by normalizing matter -element data via subordinate function and revising evaluation indexes' weights via improved AHP method to obtain improved matter -element extension model. Taking a gold mine as an example, the improved matter-element extension model was applied for goaf safety evaluation while comparing the evaluation results with a fuzzy model. Scene detection and numerical analysis was carried out for the goaf with different evaluation results from above-mentioned two models. The result of the improved matter-element extension model accords better with the actual conditions so it is reasonable and reliable to apply it to goaf safety evaluation.