Abstract: Finite element analysis has been used in a great number of deep foundation pit excavation projects to improve the efficiency and safety in construction. Its reliability depends on the accuracy of model parameters and the suitability of boundary conditions. This paper presents a model of soil parameter back analysis that is based on the artificial neural network and its application to deep foundation pit excavation. In developing this model, first we determined the structure of neutral network according to the size of monitored data and the number of parameters, and initiated the neutral network using a genetic algorithm. Then, we developed a neutral network training set via finite element calculation. After completion of all the training, we inputted the monitored data into the network and started the back calculation of soil parameters. To demonstrate the application procedure, this paper presents the results of a foundation pit excavation project, the Tianjin culture centre transport hub. Verification results show that this back analysis model is effective and accurate in calculation of the soil parameters of deep foundation pit excavation.