Determining a feasible safety factor for space trusses is an important phase in structural analysis that could have economic benefits. We know there are many kinds of imperfections in structural elements, which include both material and geometric flaws. Predicting factual behavior of structures is very difficult and occasionally impossible. Elements with initial geometric imperfections in space trusses are a common phenomenon, in addition, equivalent initial geometric imperfections can be applied for modeling of residual stresses or eccentric loading effect. The number of members in the space structures is usually high as is the diversity in the kind of initial imperfection. Therefore, there is a high likelihood that models must be analyzed. The structure must be analyzed with non-linear methods, making these approaches time consuming, and potentially uneconomical. In this study, we selected 30 cases for random analysis based on Monte Carlo methods to find the bearing capacity of the space truss. We attained results from the LUSAS program LUSAS Modeller, Version 13, UK program and these were then exported as input data to the Artificial Neural Network (ANN) program. A reasonable neural network has been found of predicting another 30 cases for load bearing capacity without any analysis and only based on the neural network program. Finally, a new approach for determining the load capacity of the space trusses was extracted and we predicted the occurrence possibility of the convenience load bearing capacity in 60 cases.