Land subsidence is a morphological phenomenon, which causes negative environmental and economic consequences for human societies. Therefore, identifying the areas prone to subsidence can be one of the necessary measures for reducing the potential risks. This study aims to evaluate the efficiency of the support vector machine (SVM) algorithm and weighted overlay index (WOI) models in zoning the rate of land subsidence hazard in Hashtgerd plain, Iran. First, the 19 criteria include groundwater depletion, groundwater extraction, aquifer thickness, alluvium thickness, aquifer recharge, well density, drainage density, groundwater depth, lithology, bedrock depth, average annual precipitation, average annual temperature, climate type, agricultural use, urban use, industrial use, distance from rivers and streams, distance from roads, distance from faults were considered. Then, the layers were weighed based on the Best-Worst Method (BWM). The results of BWM indicated that the factors of groundwater extraction (0.219), lithology (0.157), and groundwater depletion (0.079) have a greater effect on the potential for subsidence hazard. Moreover, the results of validation by performing ROC curve showed that the accuracy of RBF-SVM, LN-SVM, SIG-SVM, PL-SVM, and WOI were 95.7, 94.3, 94.9, 93.2, and 90%, respectively. Based on the ROC results, all of the models for preparing the subsidence hazard map in Hashtgerd plain exhibit excellent accuracy. Therefore, all of the models used here can predict the areas vulnerable to subsidence properly. In this study, the five land subsidence hazard maps were used as new input factors and integrated using fuzzy gamma-ensemble methods to make better hazard maps. The results of the ensemble model indicated that 19.3% of Hashtgerd plain is in the zone of high to very high sensitivity. The results of this study can help planners in managing and reducing the possible hazards of subsidence.