frequent ground collapses resulted by anthropogenic activities occur due to rapid urbanization. accurate susceptibility mapping is critical for disaster prevention and control. in this study, 1198 ground collapse cases were collected from 2017 to 2020 in shenzhen. after multicollinearity testing, eight effective factors (elevation, relief, clay proportion, average annual precipitation, distance from water, land use type, building density, and road density) were selected to construct an evaluation index system. ground collapse susceptibility was analyzed and mapped using the normalized frequency ratio (nfr), logistic regression (lr), and nfr-lr coupling models. finally, the rationality and performance of the three models were compared using the frequency ratio (fr) and receiver operating characteristic (roc) curve, respectively. all three models could effectively evaluate the ground collapse susceptibility (area under the roc curve (auc) > 0.7), though the result of nfr-lr model was more rational and had the highest performance (auc = 0.791) among the three models. shenzhen possessed ground collapses mainly in built-up areas, the greater the intensity of anthropogenic activity in a location, the higher the likelihood of a disaster. the high and very high susceptibility zones covered a total area of 546.76 km2 and involved the nanshan, luohu, and futian districts, as well as some areas of the baoan, guangming, and longgang districts.