Land subsidence poses one of the major natural hazards around the globe that cause damage to life and property. Although several advanced models have been applied to model land subsidence susceptibility, no consensus has been reached on the most accurate models to study this phenomenon. In this work, we propose the use of the following five state-of-the-art models to calculate the susceptibility to land subsidence across a region in Iran: artificial neural network – satin bowerbird optimization (ANN-SBO), artificial neural network-water cycle algorithm (ANNWCA), artificial neural network-chimp optimization algorithm (ANN-ChoA) and artificial neural network-crow search algorithm (ANN-CSA). We used 12 land subsidence predictors and 93 land subsidence locations as input data in the algorithms. The land subsidence locations were divided into training (65 locations or 70%) and validating (28 locations or 30%) samples. As per the importance factor analysis, the Groundwater Withdraw variable was found the most important factor among all input factors and the slope was found the least important factor among all. According to the validation procedure the most performing model, in terms of Success Rate, was WCA-ANN (AUC = 0.953), followed by ChOA-ANN (AUC = 0.944), SBO-ANN (AUC = 0.924), CSA-ANN (AUC = 0.915) and ANN (AUC = 0.913). For the Prediction Rate, the highest performance was achieved by WCAANN (AUC = 0.974), followed by ChOA-ANN (AUC = 0.958), SBOANN (AUC = 0.942), CSA-ANN (AUC = 0.931) and ANN (AUC = 0.927). The present work of such higher accuracy can be useful for the policymakers of govt. of Iran during operation work of any mega projects and implementation. [ABSTRACT FROM AUTHOR]