Abstract: To predict the development trend of ground settlement around foundation pit accurately, a prediction model for maximum ground settlement of foundation pit was proposed based on shuffled frog leaping algorithm and generalized regression neural network model(SFLA-GRNN model). Firstly, through the settlement mechanism analysis and the initial selection of the input variable set, grey correlation analysis was used to quantify the correlation between model input and output variables. Some of input variables that are significantly less correlated with output variables were eliminated. Secondly, the smoothing factor of the generalized regression neural network model (GRNN) was optimized by using the shuffled frog algorithm (SFLA), so as to reduce the adverse effects of human factors on the accuracy and generalization ability of the model. Finally, a generalized regression neural network model for predicting the maximum settlement of the foundation pit was established by using the selected input variables set. Example application and comparative analysis show that input variables selection based on gray correlation degree and smoothing factor optimization based on shuffled frog leaping algorithm all can effectively improve the accuracy and generalization ability of GRNN model. The above conclusions can provide reference for similar deformation prediction.

Key words: shuffled frog leaping algorithm, generalized regression neural network, smoothing factor, grey correlation analysis, settlement prediction